\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0<\/span>Review of Borehole Log Digitization Papers<\/span><\/title>\n<style>\n:root {\n --bg: #f5faf7;\n --bg2: #eef6f1;\n --ink: #1a2e23;\n --muted: #5a7a6e;\n --rule: #c0d8cc;\n --accent: #059669;\n --accent2: #0d9488;\n --code-bg: #1e2d26;\n --code-fg: #d1fae5;\n}\n\n* { box-sizing: border-box; margin: 0; padding: 0; }\n\nhtml { scroll-behavior: smooth; }\n\nbody {\n font-family: \"Microsoft YaHei\", \"PingFang SC\", system-ui, -apple-system, sans-serif;\n background: var(--bg);\n color: var(--ink);\n line-height: 1.8;\n}\n\nbody.en-active { font-family: system-ui, -apple-system, sans-serif; }\n\n\/* Language Switcher Bar *\/\n.lang-bar {\n position: sticky;\n top: 0;\n z-index: 1000;\n background: var(--accent);\n display: flex;\n justify-content: center;\n gap: 1rem;\n padding: 0.6rem 1rem;\n box-shadow: 0 2px 8px rgba(0,0,0,0.15);\n}\n\n.lang-bar button {\n background: rgba(255,255,255,0.2);\n border: 2px solid transparent;\n color: #fff;\n padding: 0.4rem 1.2rem;\n border-radius: 4px;\n cursor: pointer;\n font-size: 0.95rem;\n font-weight: 600;\n transition: all 0.2s;\n}\n\n.lang-bar button:hover {\n background: rgba(255,255,255,0.35);\n}\n\n.lang-bar button.active {\n background: #fff;\n color: var(--accent);\n border-color: #fff;\n}\n\n\/* Container *\/\n.container {\n max-width: 1100px;\n margin: 0 auto;\n padding: 0 1.5rem;\n}\n\n\/* Cover *\/\n.cover {\n text-align: center;\n padding: 4rem 1rem 3rem;\n background: var(--bg2);\n border-bottom: 1px solid var(--rule);\n}\n\n.cover h1 {\n font-size: 2.2rem;\n color: var(--ink);\n margin-bottom: 0.8rem;\n font-weight: 700;\n}\n\n.cover .subtitle {\n font-size: 1.3rem;\n color: var(--accent);\n font-weight: 600;\n margin-bottom: 0.6rem;\n}\n\n.cover .desc {\n font-size: 1rem;\n color: var(--muted);\n}\n\n\/* TOC *\/\n.toc {\n background: #fff;\n border: 1px solid var(--rule);\n border-radius: 8px;\n padding: 1.5rem 2rem;\n margin: 2rem 0;\n}\n\n.toc h2 {\n font-size: 1.3rem;\n margin-bottom: 1rem;\n color: var(--accent);\n}\n\n.toc ol, .toc ul {\n padding-left: 1.5rem;\n}\n\n.toc li {\n margin: 0.4rem 0;\n}\n\n.toc a {\n color: var(--ink);\n text-decoration: none;\n transition: color 0.2s;\n}\n\n.toc a:hover {\n color: var(--accent);\n text-decoration: underline;\n}\n\n\/* Sections *\/\nsection {\n background: #fff;\n border: 1px solid var(--rule);\n border-radius: 8px;\n padding: 2rem 2.5rem;\n margin: 2rem 0;\n}\n\nsection h2 {\n font-size: 1.6rem;\n color: var(--accent);\n margin-bottom: 1.2rem;\n padding-bottom: 0.5rem;\n border-bottom: 2px solid var(--rule);\n}\n\nsection h3 {\n font-size: 1.25rem;\n color: var(--accent2);\n margin: 1.5rem 0 0.8rem;\n}\n\nsection h4 {\n font-size: 1.1rem;\n color: var(--ink);\n margin: 1.2rem 0 0.6rem;\n font-weight: 600;\n}\n\nsection p {\n margin-bottom: 0.9rem;\n text-align: justify;\n}\n\n\/* Tables *\/\ntable {\n width: 100%;\n border-collapse: collapse;\n margin: 1.2rem 0;\n font-size: 0.95rem;\n}\n\nth, td {\n border: 1px solid var(--rule);\n padding: 0.7rem 0.9rem;\n text-align: left;\n}\n\nth {\n background: var(--bg2);\n font-weight: 600;\n color: var(--ink);\n}\n\ntr:nth-child(even) { background: var(--bg); }\n\n\/* Code *\/\npre {\n background: var(--code-bg);\n color: var(--code-fg);\n padding: 1.2rem 1.5rem;\n border-radius: 6px;\n overflow-x: auto;\n font-family: Consolas, \"Courier New\", monospace;\n font-size: 0.9rem;\n line-height: 1.6;\n margin: 1rem 0;\n}\n\ncode {\n font-family: Consolas, \"Courier New\", monospace;\n background: rgba(5,150,105,0.1);\n padding: 0.15rem 0.4rem;\n border-radius: 3px;\n font-size: 0.9em;\n color: var(--accent);\n}\n\npre code {\n background: transparent;\n padding: 0;\n color: inherit;\n}\n\n\/* Tip box *\/\n.tip-box {\n background: #ecfdf5;\n border-left: 4px solid var(--accent);\n padding: 1rem 1.2rem;\n margin: 1.2rem 0;\n border-radius: 0 6px 6px 0;\n font-size: 0.95rem;\n}\n\n.tip-box.warn {\n background: #fffbeb;\n border-left-color: #f59e0b;\n}\n\n.tip-box.info {\n background: #eff6ff;\n border-left-color: #3b82f6;\n}\n\n.tip-box strong { color: var(--accent); }\n.tip-box.warn strong { color: #b45309; }\n.tip-box.info strong { color: #1d4ed8; }\n\n\/* Highlight *\/\nmark.key {\n background: #d1fae5;\n color: #065f46;\n padding: 0.15em 0.4em;\n border-radius: 3px;\n font-weight: 700;\n}\n\n\/* Lists *\/\nul, ol {\n padding-left: 1.8rem;\n margin: 0.8rem 0;\n}\n\nli { margin: 0.4rem 0; }\n\n\/* Footnotes *\/\n.footnotes {\n font-size: 0.85rem;\n color: var(--muted);\n border-top: 1px solid var(--rule);\n margin-top: 2rem;\n padding-top: 1rem;\n text-align: center;\n}\n\n\/* Table wrap for scrolling *\/\n.table-wrap {\n overflow-x: auto;\n margin-bottom: 1rem;\n}\n\n\/* Section icon *\/\n.section-icon {\n width: 32px;\n height: 32px;\n border-radius: 8px;\n background: var(--bg2);\n display: inline-flex;\n align-items: center;\n justify-content: center;\n font-size: 1rem;\n margin-right: 0.6rem;\n vertical-align: middle;\n}\n\n\/* Paper card *\/\n.paper-card {\n background: var(--bg2);\n border: 1px solid var(--rule);\n border-radius: 8px;\n padding: 1.5rem;\n margin: 1.2rem 0;\n}\n\n.paper-title {\n font-size: 1.15rem;\n font-weight: 700;\n color: var(--ink);\n margin-bottom: 0.5rem;\n}\n\n.paper-meta {\n font-size: 0.9rem;\n color: var(--muted);\n margin-bottom: 1rem;\n}\n\n\/* Method flow *\/\n.method-flow {\n display: flex;\n align-items: center;\n flex-wrap: wrap;\n gap: 0.5rem;\n margin: 1rem 0;\n padding: 1rem;\n background: var(--bg);\n border-radius: 6px;\n}\n\n.method-step {\n background: #fff;\n border: 1px solid var(--rule);\n padding: 0.4rem 0.8rem;\n border-radius: 4px;\n font-size: 0.9rem;\n font-weight: 600;\n}\n\n.method-arrow {\n color: var(--accent);\n font-weight: bold;\n}\n\n\/* Innovation tags *\/\n.innovation-tag {\n display: inline-block;\n background: #d1fae5;\n color: #065f46;\n padding: 0.2em 0.6em;\n border-radius: 4px;\n font-size: 0.8rem;\n font-weight: 600;\n margin-right: 0.4rem;\n margin-bottom: 0.3rem;\n}\n\n\/* Responsive *\/\n@media (max-width: 768px) {\n .cover h1 { font-size: 1.5rem; }\n .cover .subtitle { font-size: 1.05rem; }\n section { padding: 1.2rem 1rem; }\n section h2 { font-size: 1.3rem; }\n table { font-size: 0.85rem; }\n th, td { padding: 0.5rem; }\n .method-flow { flex-direction: column; align-items: flex-start; }\n}\n\n\/* Language display control *\/\n[data-lang=\"en\"] { display: none; }\nhtml[lang=\"en\"] [data-lang=\"zh\"] { display: none; }\nhtml[lang=\"en\"] [data-lang=\"en\"] { display: inline; }\n\nh2 [data-lang=\"en\"], h3 [data-lang=\"en\"], h4 [data-lang=\"en\"], p [data-lang=\"en\"], li [data-lang=\"en\"], th [data-lang=\"en\"], td [data-lang=\"en\"], div [data-lang=\"en\"], strong [data-lang=\"en\"] {\n display: none;\n}\n\nhtml[lang=\"en\"] h2 [data-lang=\"zh\"], html[lang=\"en\"] h3 [data-lang=\"zh\"], html[lang=\"en\"] h4 [data-lang=\"zh\"], html[lang=\"en\"] p [data-lang=\"zh\"], html[lang=\"en\"] li [data-lang=\"zh\"], html[lang=\"en\"] th [data-lang=\"zh\"], html[lang=\"en\"] td [data-lang=\"zh\"], html[lang=\"en\"] div [data-lang=\"zh\"], html[lang=\"en\"] strong [data-lang=\"zh\"] {\n display: none;\n}\n\nhtml[lang=\"en\"] h2 [data-lang=\"en\"], html[lang=\"en\"] h3 [data-lang=\"en\"], html[lang=\"en\"] h4 [data-lang=\"en\"], html[lang=\"en\"] p [data-lang=\"en\"], html[lang=\"en\"] li [data-lang=\"en\"], html[lang=\"en\"] th [data-lang=\"en\"], html[lang=\"en\"] td [data-lang=\"en\"], html[lang=\"en\"] div [data-lang=\"en\"], html[lang=\"en\"] strong [data-lang=\"en\"] {\n display: inline;\n}\n\n.toc h2 span, .cover h1 span, .cover .subtitle span, .cover .desc span, section h2 span, section h3 span, section h4 span {\n display: inline;\n}\n\nhtml[lang=\"en\"] .toc h2 [data-lang=\"zh\"], html[lang=\"en\"] .cover h1 [data-lang=\"zh\"], html[lang=\"en\"] .cover .subtitle [data-lang=\"zh\"], html[lang=\"en\"] .cover .desc [data-lang=\"zh\"], html[lang=\"en\"] section h2 [data-lang=\"zh\"], html[lang=\"en\"] section h3 [data-lang=\"zh\"], html[lang=\"en\"] section h4 [data-lang=\"zh\"] {\n display: none;\n}\n\nhtml[lang=\"en\"] .toc h2 [data-lang=\"en\"], html[lang=\"en\"] .cover h1 [data-lang=\"en\"], html[lang=\"en\"] .cover .subtitle [data-lang=\"en\"], html[lang=\"en\"] .cover .desc [data-lang=\"en\"], html[lang=\"en\"] section h2 [data-lang=\"en\"], html[lang=\"en\"] section h3 [data-lang=\"en\"], html[lang=\"en\"] section h4 [data-lang=\"en\"] {\n display: inline;\n}\n<\/style>\n<\/head>\n<body>\n\n\n<div class=\"lang-bar\">\n <button id=\"btn-zh\" class=\"active\" onclick=\"switchLanguage('zh')\">\n \u4e2d\u6587<\/span>Chinese<\/span>\n <\/button>\n <button id=\"btn-en\" onclick=\"switchLanguage('en')\">\n English<\/span>English<\/span>\n <\/button>\n<\/div>\n\n<div class=\"container\">\n\n\n<div class=\"cover\">\n <h1>\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0<\/span>Review of Borehole Log Digitization Papers<\/span><\/h1>\n <div class=\"subtitle\">Academic Research Review<\/div>\n <div class=\"desc\">\u7cfb\u7edf\u68b3\u7406\u94bb\u5b54\u6570\u636e\u6570\u5b57\u5316\u3001\u5730\u5c42\u89e3\u91ca\u3001OCR\u8bc6\u522b\u7b49\u9886\u57df\u6838\u5fc3\u8bba\u6587\u7684\u65b9\u6cd5\u8bba\u4e0e\u521b\u65b0\u8d21\u732e\u3002<\/span>Systematic review of methodologies and innovations in borehole data digitization, stratigraphic interpretation, and OCR recognition.<\/span><\/div>\n<\/div>\n\n\n<nav class=\"toc\">\n <h2>\u76ee\u5f55<\/span>Table of Contents<\/span><\/h2>\n <ol>\n <li><a href=\"#overview\">\u7814\u7a76\u8109\u7edc\u6982\u8ff0<\/span>Research Overview<\/span><\/a><\/li>\n <li><a href=\"#international\">\u56fd\u9645\u671f\u520a\u8bba\u6587<\/span>International Journal Papers<\/span><\/a><\/li>\n <li><a href=\"#chinese\">\u4e2d\u6587\u6838\u5fc3\u7814\u7a76<\/span>Chinese Core Research<\/span><\/a><\/li>\n <li><a href=\"#conference\">\u4f1a\u8bae\u8bba\u6587<\/span>Conference Papers<\/span><\/a><\/li>\n <li><a href=\"#comparison\">\u65b9\u6cd5\u5bf9\u6bd4\u4e0e\u521b\u65b0\u603b\u7ed3<\/span>Method Comparison & Innovation Summary<\/span><\/a><\/li>\n <\/ol>\n<\/nav>\n\n<section class=\"section\" id=\"overview\">\n\n<div class=\"section-icon\">\ud83d\udcd6<\/div>\n<h2>\u7814\u7a76\u8109\u7edc\u6982\u8ff0<\/span>Research Overview<\/span><\/h2>\n\n\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u76f8\u5173\u7684\u5b66\u672f\u7814\u7a76\u53ef\u5206\u4e3a\u56db\u5927\u4e3b\u7ebf\uff1a<\/span>\nAcademic research on borehole log digitization can be divided into four main threads:<\/span><\/p>\n\n<ol>\n<li>\u81ea\u52a8\u5316\u6570\u636e\u63d0\u53d6<\/span>Automated Data Extraction<\/span><\/strong> \u2014 \n\u4ece\u975e\u7ed3\u6784\u5316\u5730\u8d28\u62a5\u544a(PDF\/\u626b\u63cf\u4ef6)\u4e2d\u81ea\u52a8\u63d0\u53d6\u5b54\u53e3\u3001\u6d4b\u659c\u3001\u5ca9\u6027\u7b49\u7ed3\u6784\u5316\u6570\u636e<\/span>\nAutomatically extracting structured data (collar, survey, lithology) from unstructured geological reports (PDF\/scans)<\/span><\/li>\n<li>\u5730\u5c42\u5e8f\u5217\u6062\u590d\u4e0e\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Stratigraphic Sequence Recovery & Uncertainty Quantification<\/span><\/strong> \u2014 \n\u4ece\u5ca9\u6027\u63cf\u8ff0\u4e2d\u6062\u590d\u5730\u5c42\u5e8f\u5217\uff0c\u8bc4\u4f30\u89e3\u91ca\u7ed3\u679c\u7684\u4e0d\u786e\u5b9a\u6027<\/span>\nRecovering stratigraphic sequences from lithological descriptions and quantifying interpretation uncertainty<\/span><\/li>\n<li>OCR\u4e0e\u56fe\u50cf\u8bc6\u522b\u4f18\u5316<\/span>OCR & Image Recognition Optimization<\/span><\/strong> \u2014 \n\u63d0\u5347\u94bb\u5b54\u67f1\u72b6\u56fe\u56fe\u50cf\u4e2d\u6587\u5b57\u548c\u7b26\u53f7\u7684\u8bc6\u522b\u51c6\u786e\u7387<\/span>\nImproving recognition accuracy of text and symbols in borehole log images<\/span><\/li>\n<li>\u6570\u636e\u5e93\u5efa\u8bbe\u4e0e\u8de8\u5e73\u53f0\u5e94\u7528<\/span>Database Construction & Cross-Platform Applications<\/span><\/strong> \u2014 \n\u5efa\u7acb\u6807\u51c6\u5316\u7684\u94bb\u5b54\u6570\u636e\u5e93\uff0c\u5b9e\u73b0\u6570\u636e\u5171\u4eab\u548c\u8de8\u5e73\u53f0\u53ef\u89c6\u5316<\/span>\nEstablishing standardized borehole databases for data sharing and cross-platform visualization<\/span><\/li>\n<\/ol>\n\n<\/section>\n\n<section class=\"section\" id=\"international\">\n\n<div class=\"section-icon\">\ud83c\udf0d<\/div>\n<h2>\u56fd\u9645\u671f\u520a\u8bba\u6587<\/span>International Journal Papers<\/span><\/h2>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">dh2loop 1.0: an open-source python library for automated processing and classification of geological logs<\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1aGeoscientific Model Development (2026) | \u4f5c\u8005\uff1aRanee Joshi \u7b49<\/span>\nJournal: Geoscientific Model Development (2026) | Authors: Ranee Joshi et al.<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u5730\u8d28\u94bb\u5b54\u62a5\u544a\u4ee5\u975e\u7ed3\u6784\u5316\u6587\u672c\u5f62\u5f0f\u5b58\u50a8\u4e8e\u5404\u5730\u8d28\u8c03\u67e5\u5c40\u6570\u636e\u5e93\u4e2d\uff0c\u5ca9\u6027\u63cf\u8ff0\u5b58\u5728\u4e25\u91cd\u7684\u547d\u540d\u4e0d\u4e00\u81f4\u95ee\u9898\uff08\u540c\u4e00\u5ca9\u6027\u53ef\u80fd\u6709\u6570\u5341\u79cd\u4e0d\u540c\u63cf\u8ff0\u65b9\u5f0f\uff09\uff0c\u963b\u788d\u4e86\u5927\u89c4\u6a21\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u7684\u6570\u636e\u51c6\u5907\u3002<\/span>\nGeological drillhole reports are stored as unstructured text in geological survey databases. Lithological descriptions suffer from severe naming inconsistency (the same lithology may have dozens of different descriptions), hindering data preparation for large-scale 3D geological modeling.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<div class=\"method-flow\">\nPDF\u89e3\u6790<\/span>PDF Parsing<\/span><\/span>\n\u2192<\/span>\n\u6587\u672c\u5206\u5757<\/span>Text Chunking<\/span><\/span>\n\u2192<\/span>\n\u6b63\u5219\u5339\u914d<\/span>Regex Matching<\/span><\/span>\n\u2192<\/span>\n\u8bcd\u5e93\u5339\u914d<\/span>Lexicon Matching<\/span><\/span>\n\u2192<\/span>\n\u6a21\u7cca\u5b57\u7b26\u4e32\u5339\u914d<\/span>Fuzzy Matching<\/span><\/span>\n\u2192<\/span>\n\u591a\u7ea7\u5206\u7c7b<\/span>Multi-level Classification<\/span><\/span>\n<\/div>\n\n<ul>\n<li>\u4e09\u6bb5\u5f0f\u6570\u636e\u6a21\u578b\uff1a<\/strong>\u5c06\u94bb\u5b54\u6570\u636e\u6807\u51c6\u5316\u4e3a\u5b54\u53e3(collar)\u3001\u6d4b\u659c(survey)\u3001\u5ca9\u6027(lithology)\u4e09\u4e2a\u6838\u5fc3\u95f4\u9694\u8868<\/span>\nThree-part Data Model:<\/strong> Standardizes drillhole data into collar, survey, and lithology interval tables<\/span><\/li>\n<li>\u8bcd\u5e93\u9a71\u52a8\u5339\u914d\uff1a<\/strong>\u6784\u5efa\u6807\u51c6\u5316\u5ca9\u6027\u8bcd\u5e93\uff0c\u5c06\u539f\u59cb\u63cf\u8ff0\u6620\u5c04\u5230\u6807\u51c6\u672f\u8bed<\/span>\nLexicon-driven Matching:<\/strong> Builds standardized lithology lexicon to map raw descriptions to standard terms<\/span><\/li>\n<li>\u6a21\u7cca\u5339\u914d\u7b97\u6cd5\uff1a<\/strong>\u91c7\u7528\u6a21\u7cca\u5b57\u7b26\u4e32\u5339\u914d(fuzzy string matching)\u5904\u7406\u62fc\u5199\u53d8\u4f53\u548c\u540c\u4e49\u8bcd<\/span>\nFuzzy Matching:<\/strong> Uses fuzzy string matching to handle spelling variants and synonyms<\/span><\/li>\n<li>\u591a\u7ea7\u5206\u7c7b\u4f53\u7cfb\uff1a<\/strong>\u652f\u6301\u4ece\u8be6\u7ec6\u5ca9\u6027\u5230\u7c97\u7c92\u5ea6\u7c7b\u522b\u7684\u591a\u7ea7\u5f52\u5e76\uff0c\u9002\u914d\u4e0d\u540c\u5c3a\u5ea6\u7684\u4e09\u7ef4\u5efa\u6a21\u9700\u6c42<\/span>\nMulti-level Taxonomy:<\/strong> Supports aggregation from detailed lithology to coarse categories for different modeling scales<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u5f00\u6e90\u5b9e\u73b0<\/span>Open Source<\/span><\/span>\n\u8bcd\u5e93\u6807\u51c6\u5316<\/span>Lexicon Standardization<\/span><\/span>\n\u591a\u7ea7\u5206\u7c7b<\/span>Hierarchical Classification<\/span><\/span>\n\n<ul>\n<li>\u9996\u6b21\u5c06\u5b8c\u6574\u7684\u94bb\u5b54\u6570\u636e\u63d0\u53d6\u6d41\u7a0b\u5f00\u6e90\u4e3a\u53ef\u590d\u7528Python\u5e93<\/span>\nFirst open-source Python library providing a complete drillhole data extraction pipeline<\/span><\/li>\n<li>\u63d0\u51fa\u8bcd\u5e93+\u6a21\u7cca\u5339\u914d\u7684\u4e24\u9636\u6bb5\u5ca9\u6027\u547d\u540d\u6d88\u6b67\u65b9\u6cd5<\/span>\nProposes a two-stage lithology name disambiguation method: lexicon + fuzzy matching<\/span><\/li>\n<li>\u591a\u7ea7\u5ca9\u6027\u5206\u7c7b\u673a\u5236\u89e3\u51b3\u4e86\u4ece\u533a\u57df\u5efa\u6a21\u5230\u5c40\u90e8\u5efa\u6a21\u7684\u6570\u636e\u7c92\u5ea6\u9002\u914d\u95ee\u9898<\/span>\nMulti-level classification mechanism solves data granularity adaptation from regional to local modeling<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box\">\n\u5b9e\u9a8c\u7ed3\u679c\uff1a<\/span>Experimental Results:<\/span><\/strong><br\/>\n\u5b54\u53e3\u6570\u636e\u63d0\u53d6\u7387 93%<\/mark>\uff0c\u6d4b\u659c\u6570\u636e 86%<\/mark>\uff0c\u5ca9\u6027\u6570\u636e 34%<\/mark>\uff1b\u5176\u4e2d86%\u7684\u5ca9\u6027\u6570\u636e\u53ef\u6210\u529f\u5339\u914d\u5230\u6807\u51c6\u8bcd\u5e93\u3002<\/span>\nCollar extraction rate 93%<\/mark>, survey 86%<\/mark>, lithology 34%<\/mark>; 86% of lithology data successfully matched to standard lexicon.<\/span>\n<\/div>\n<\/div>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">Automated stratigraphic interpretation from drillhole lithological descriptions with uncertainty quantification: litho2strat 1.0<\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1aGeoscientific Model Development (2026)<\/span>\nJournal: Geoscientific Model Development (2026)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u4ece\u94bb\u5b54\u5ca9\u6027\u6570\u636e\u6062\u590d\u5730\u5c42\u5e8f\u5217\u662f\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u7684\u5173\u952e\u6b65\u9aa4\uff0c\u4f46\u4f20\u7edf\u4eba\u5de5\u89e3\u91ca\u6548\u7387\u4f4e\u4e14\u4e3b\u89c2\u6027\u5f3a\u3002\u5982\u4f55\u81ea\u52a8\u4ece\u5ca9\u6027\u63cf\u8ff0\u4e2d\u63a8\u65ad\u5730\u5c42\u5e8f\u5217\uff0c\u5e76\u91cf\u5316\u89e3\u91ca\u7684\u4e0d\u786e\u5b9a\u6027\uff0c\u662f\u6838\u5fc3\u79d1\u5b66\u95ee\u9898\u3002<\/span>\nRecovering stratigraphic sequences from drillhole lithology data is a key step in 3D geological modeling, but traditional manual interpretation is inefficient and subjective. Automatically inferring stratigraphic sequences from lithological descriptions while quantifying uncertainty is the core scientific problem.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<div class=\"method-flow\">\n\u5ca9\u6027\u8f93\u5165<\/span>Lithology Input<\/span><\/span>\n\u2192<\/span>\n\u641c\u7d22\u7b97\u6cd5<\/span>Search Algorithm<\/span><\/span>\n\u2192<\/span>\n\u5730\u5c42\u6392\u5e8f<\/span>Stratigraphic Ordering<\/span><\/span>\n\u2192<\/span>\n\u8de8\u5b54\u5bf9\u6bd4<\/span>Cross-hole Correlation<\/span><\/span>\n\u2192<\/span>\n\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Uncertainty Quantification<\/span><\/span>\n<\/div>\n\n<ul>\n<li>\u7cfb\u7edf\u641c\u7d22\u7b97\u6cd5\uff1a<\/strong>\u901a\u8fc7\u641c\u7d22\u7b97\u6cd5\u7cfb\u7edf\u63a2\u7d22\u6240\u6709\u5730\u8d28\u5408\u7406\u7684\u5730\u5c42\u6392\u5e8f\u65b9\u6848\uff0c\u800c\u975e\u5355\u4e00\u786e\u5b9a\u6027\u89e3<\/span>\nSystematic Search:<\/strong> Explores all geologically plausible stratigraphic orderings via search algorithms, not a single deterministic solution<\/span><\/li>\n<li>\u5730\u8d28\u5408\u7406\u6027\u7ea6\u675f\uff1a<\/strong>\u5f15\u5165\u5730\u8d28\u89c4\u5219\u7ea6\u675f\u641c\u7d22\u7a7a\u95f4\uff0c\u5982\u5730\u5c42\u53e0\u8986\u5f8b\u3001\u5ca9\u6027\u8fc7\u6e21\u6982\u7387\u7b49<\/span>\nGeological Plausibility Constraints:<\/strong> Introduces geological rules (superposition law, lithology transition probabilities) to constrain search space<\/span><\/li>\n<li>\u8de8\u94bb\u5b54\u5bf9\u6bd4\uff1a<\/strong>\u7ed3\u5408\u591a\u94bb\u5b54\u95f4\u5730\u5c42\u5355\u5143\u7684\u62d3\u6251\u5173\u7cfb\u8fdb\u884c\u5bf9\u6bd4\u9a8c\u8bc1\uff0c\u589e\u5f3a\u89e3\u91ca\u53ef\u9760\u6027<\/span>\nCross-hole Correlation:<\/strong> Correlates stratigraphic units across multiple drillholes using topological relationships to enhance reliability<\/span><\/li>\n<li>\u4e0d\u786e\u5b9a\u6027\u91cf\u5316\uff1a<\/strong>\u8f93\u51fa\u591a\u4e2a\u5730\u8d28\u4e00\u81f4\u6027\u89e3\u53ca\u5176\u6982\u7387\u5206\u5e03\uff0c\u4e3a\u5efa\u6a21\u63d0\u4f9b\u98ce\u9669\u4f9d\u636e<\/span>\nUncertainty Quantification:<\/strong> Outputs multiple geologically consistent solutions with probability distributions, providing risk basis for modeling<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Uncertainty Quantification<\/span><\/span>\n\u8de8\u5b54\u62d3\u6251<\/span>Cross-hole Topology<\/span><\/span>\nOpen Source<\/span>\u5f00\u6e90\u5b9e\u73b0<\/span><\/span>\n\n<ul>\n<li>\u5c06\u5730\u5c42\u89e3\u91ca\u4ece\u786e\u5b9a\u6027\u95ee\u9898\u91cd\u65b0\u5b9a\u4e49\u4e3a\u6982\u7387\u63a8\u65ad\u95ee\u9898\uff0c\u9996\u6b21\u5b9e\u73b0\u5168\u81ea\u52a8\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>\nRedefines stratigraphic interpretation from a deterministic to a probabilistic inference problem, achieving fully automated uncertainty quantification<\/span><\/li>\n<li>\u63d0\u51fa\u8de8\u94bb\u5b54\u5730\u5c42\u5355\u5143\u62d3\u6251\u5173\u7cfb\u5bf9\u6bd4\u65b9\u6cd5\uff0c\u5c06\u5355\u5b54\u89e3\u91ca\u7684\u53ef\u9760\u6027\u63d0\u5347\u5230\u591a\u5b54\u8054\u5408\u7ea6\u675f\u5c42\u9762<\/span>\nProposes cross-hole stratigraphic unit topological correlation, elevating single-hole interpretation reliability to multi-hole joint constraint level<\/span><\/li>\n<li>\u5f00\u6e90\u7b97\u6cd5litho2strat 1.0\u53ef\u76f4\u63a5\u96c6\u6210\u5230\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u5de5\u4f5c\u6d41\u4e2d<\/span>\nOpen-source algorithm litho2strat 1.0 can be directly integrated into 3D geological modeling workflows<\/span><\/li>\n<\/ul>\n<\/div>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">Optimizing Text Recognition in Borehole Log Images Using a Multi-Layout Adjustment Voting Mechanism<\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1aApplied Sciences (MDPI, 2025)<\/span>\nJournal: Applied Sciences (MDPI, 2025)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u94bb\u5b54\u67f1\u72b6\u56fe\u56fe\u50cf\u7684\u6587\u5b57\u8bc6\u522b\u51c6\u786e\u7387\u53d7\u9650\u4e8e\u590d\u6742\u5e03\u5c40\uff08\u591a\u680f\u3001\u65cb\u8f6c\u6587\u5b57\u3001\u8868\u683c\u7ebf\u5e72\u6270\u7b49\uff09\uff0c\u4f20\u7edfOCR\u76f4\u63a5\u5e94\u7528\u6548\u679c\u4e0d\u4f73\u3002<\/span>\nText recognition accuracy in borehole log images is limited by complex layouts (multi-column, rotated text, table line interference, etc.), where traditional OCR performs poorly.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<div class=\"method-flow\">\n\u56fe\u50cf\u8f93\u5165<\/span>Image Input<\/span><\/span>\n\u2192<\/span>\n\u591a\u5e03\u5c40\u8c03\u6574<\/span>Multi-layout Adjustment<\/span><\/span>\n\u2192<\/span>\n\u591a\u6b21OCR<\/span>Multiple OCR Runs<\/span><\/span>\n\u2192<\/span>\n\u6295\u7968\u6574\u5408<\/span>Voting Integration<\/span><\/span>\n\u2192<\/span>\n\u6700\u4f18\u7ed3\u679c<\/span>Optimal Result<\/span><\/span>\n<\/div>\n\n<ul>\n<li>\u591a\u5e03\u5c40\u8c03\u6574\u7b56\u7565\uff1a<\/strong>\u5bf9\u539f\u59cb\u56fe\u50cf\u8fdb\u884c\u591a\u79cd\u5e03\u5c40\u53d8\u6362\uff08\u5982\u5206\u680f\u5207\u5272\u3001\u65cb\u8f6c\u6821\u6b63\u3001\u5bf9\u6bd4\u5ea6\u589e\u5f3a\u7b49\uff09\uff0c\u751f\u6210\u591a\u4e2a\u5019\u9009\u8f93\u5165<\/span>\nMulti-layout Adjustment:<\/strong> Applies various layout transformations (column splitting, rotation correction, contrast enhancement) to generate multiple candidate inputs<\/span><\/li>\n<li>\u591aOCR\u7ed3\u679c\u751f\u6210\uff1a<\/strong>\u6bcf\u4e2a\u5019\u9009\u5e03\u5c40\u72ec\u7acb\u8fd0\u884cOCR\uff0c\u4ea7\u751f\u591a\u4e2a\u8bc6\u522b\u7ed3\u679c<\/span>\nMultiple OCR Results:<\/strong> Each candidate layout runs OCR independently, producing multiple recognition results<\/span><\/li>\n<li>\u6295\u7968\u673a\u5236\u6574\u5408\uff1a<\/strong>\u8bbe\u8ba1\u6295\u7968\u7b97\u6cd5\u5bf9\u591a\u8defOCR\u7ed3\u679c\u8fdb\u884c\u7f6e\u4fe1\u5ea6\u52a0\u6743\u6574\u5408\uff0c\u8f93\u51fa\u6700\u4f18\u8bc6\u522b\u7ed3\u679c<\/span>\nVoting Integration:<\/strong> Designed voting algorithm performs confidence-weighted integration of multi-path OCR results to output optimal recognition<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u591a\u5e03\u5c40\u6295\u7968<\/span>Multi-layout Voting<\/span><\/span>\n\u5de5\u7a0b\u6570\u636e\u96c6<\/span>Engineering Dataset<\/span><\/span>\n\n<ul>\n<li>\u63d0\u51fa”\u591a\u5e03\u5c40\u8c03\u6574+\u6295\u7968”\u7684\u65b0\u8303\u5f0f\uff0c\u7a81\u7834\u4e86\u5355\u4e00\u8def\u5f84OCR\u7684\u6027\u80fd\u74f6\u9888<\/span>\nProposes a “multi-layout adjustment + voting” paradigm, breaking through the performance bottleneck of single-path OCR<\/span><\/li>\n<li>\u5728\u5927\u5c39\u683c\u5e84\u548c\u4e01\u5bb6\u5c71\u4e24\u4e2a\u771f\u5b9e\u5730\u8d28\u6570\u636e\u96c6\u4e0a\u9a8c\u8bc1\uff0cF1\u503c\u5206\u522b\u8fbe\u523097.96%\u548c94.36%<\/span>\nValidated on two real geological datasets (Dayingezhuang and Dingjiashan), achieving F1 scores of 97.96% and 94.36% respectively<\/span><\/li>\n<li>\u65b9\u6cd5\u901a\u7528\u6027\u5f3a\uff0c\u53ef\u63a8\u5e7f\u5230\u5176\u4ed6\u590d\u6742\u8868\u683c\u6587\u6863\u7684OCR\u573a\u666f<\/span>\nMethod is generalizable and can be extended to other complex table document OCR scenarios<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box success\">\n\u5b9e\u9a8c\u7ed3\u679c\uff1a<\/span>Experimental Results:<\/span><\/strong><br\/>\n\u5927\u5c39\u683c\u5e84\u6570\u636e\u96c6 F1 = 97.96%<\/mark> | \u4e01\u5bb6\u5c71\u6570\u636e\u96c6 F1 = 94.36%<\/mark><\/span>\nDayingezhuang F1 = 97.96%<\/mark> | Dingjiashan F1 = 94.36%<\/mark><\/span>\n<\/div>\n<\/div>\n\n<\/section>\n\n<section class=\"section\" id=\"chinese\">\n\n<div class=\"section-icon\">\ud83c\udde8\ud83c\uddf3<\/div>\n<h2>\u4e2d\u6587\u6838\u5fc3\u7814\u7a76<\/span>Chinese Core Research<\/span><\/h2>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">\u6cb3\u5317\u77011\u223650 000\u6c99\u6d41\u6cb3\u5e45\u7b2c\u56db\u7cfb\u94bb\u5b54\u6570\u636e\u5e93<\/span>Quaternary Drilling Database of Shaliuhe Sheet, Hebei Province (1:50,000)<\/span><\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1a\u300a\u4e2d\u56fd\u5730\u8d28\u300b(2020)<\/span>\nJournal: Geology in China (2020)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u533a\u57df\u5730\u8d28\u8c03\u67e5\u4ea7\u751f\u5927\u91cf\u94bb\u5b54\u6570\u636e\uff0c\u4f46\u7f3a\u4e4f\u7cfb\u7edf\u5316\u7684\u6570\u636e\u5e93\u5efa\u8bbe\u65b9\u6cd5\uff0c\u6570\u636e\u8d28\u91cf\u53c2\u5dee\u4e0d\u9f50\uff0c\u96be\u4ee5\u652f\u6491\u540e\u7eed\u7684\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u548c\u8d44\u6e90\u5171\u4eab\u3002<\/span>\nRegional geological surveys produce large amounts of drillhole data, but systematic database construction methods are lacking, data quality is inconsistent, and supporting 3D geological modeling and resource sharing is difficult.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<ul>\n<li>\u591a\u5c42\u6570\u636e\u6a21\u578b\uff1a<\/strong>\u5efa\u7acb1\u4e2a\u94bb\u5b54\u57fa\u672c\u4fe1\u606f\u56fe\u5c42 + 5\u4e2a\u94bb\u5b54\u7f16\u5f55\u6570\u636e\u5e93 + 5\u4e2a\u94bb\u5b54\u8bbe\u8ba1\u6570\u636e\u5e93 + 5\u5f20\u94bb\u5b54\u67f1\u72b6\u56fe\u7684\u5206\u5c42\u7ec4\u7ec7\u7ed3\u6784<\/span>\nMulti-layer Data Model:<\/strong> Establishes a hierarchical structure: 1 basic borehole info layer + 5 drilling log databases + 5 drilling design databases + 5 borehole columnar diagrams<\/span><\/li>\n<li>\u6570\u636e\u8d28\u91cf\u63a7\u5236\u4f53\u7cfb\uff1a<\/strong>\u4ece\u6570\u636e\u6765\u6e90\u3001\u5904\u7406\u65b9\u6cd5\u3001\u5c5e\u6027\u7ed3\u6784\u4e09\u4e2a\u7ef4\u5ea6\u5efa\u7acb\u8d28\u91cf\u63a7\u5236\u6d41\u7a0b<\/span>\nData Quality Control System:<\/strong> Establishes QC processes from three dimensions: data source, processing method, and attribute structure<\/span><\/li>\n<li>\u6807\u51c6\u5316\u5c5e\u6027\u7ed3\u6784\uff1a<\/strong>\u7edf\u4e00\u94bb\u5b54\u7f16\u53f7\u3001\u5750\u6807\u3001\u5b54\u53e3\u6807\u9ad8\u3001\u94bb\u673a\u7c7b\u578b\u3001\u5f00\u7ec8\u5b54\u65e5\u671f\u7b49\u5b57\u6bb5\u89c4\u8303<\/span>\nStandardized Attribute Structure:<\/strong> Unifies field specifications for borehole ID, coordinates, collar elevation, rig type, start\/end dates, etc.<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u6570\u636e\u5e93\u4f53\u7cfb<\/span>Database System<\/span><\/span>\n\u8d28\u63a7\u6d41\u7a0b<\/span>QC Process<\/span><\/span>\n\n<ul>\n<li>\u9996\u6b21\u7cfb\u7edf\u9610\u8ff0\u4e861:5\u4e07\u6bd4\u4f8b\u5c3a\u7b2c\u56db\u7cfb\u94bb\u5b54\u6570\u636e\u5e93\u7684\u5b8c\u6574\u5efa\u8bbe\u65b9\u6cd5<\/span>\nFirst systematic description of complete construction methods for 1:50,000 scale Quaternary borehole databases<\/span><\/li>\n<li>\u63d0\u51fa\u7684\u5206\u5c42\u6570\u636e\u7ec4\u7ec7\u7ed3\u6784\u517c\u987e\u4e86\u539f\u59cb\u7f16\u5f55\u6570\u636e\u3001\u8bbe\u8ba1\u6570\u636e\u548c\u6210\u679c\u56fe\u4ef6\u7684\u7ba1\u7406\u9700\u6c42<\/span>\nThe proposed hierarchical data organization accommodates management needs for original log data, design data, and output maps<\/span><\/li>\n<\/ul>\n<\/div>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">\u8de8\u5e73\u53f0\u94bb\u5b54\u67f1\u72b6\u56fe\u7684\u81ea\u52a8\u751f\u6210<\/span>Cross-Platform Automatic Generation of Borehole Columnar Diagrams<\/span><\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1a\u300a\u5ca9\u571f\u5de5\u7a0b\u6280\u672f\u300b(2016)<\/span>\nJournal: Geotechnical Engineering Technique (2016)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u4f20\u7edf\u94bb\u5b54\u67f1\u72b6\u56fe\u4e0e\u7279\u5b9aCAD\u5e73\u53f0\u5f3a\u7ed1\u5b9a\uff0c\u4e0d\u540c\u5355\u4f4d\u4f7f\u7528\u4e0d\u540c\u8f6f\u4ef6\u5bfc\u81f4\u56fe\u4ef6\u65e0\u6cd5\u4e92\u901a\uff0c\u6570\u636e\u590d\u7528\u56f0\u96be\u3002<\/span>\nTraditional borehole columnar diagrams are tightly bound to specific CAD platforms. Different organizations using different software prevents interoperability and data reuse.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<ul>\n<li>\u901a\u7528\u6570\u636e\u6a21\u578b\uff1a<\/strong>\u8bbe\u8ba1\u72ec\u7acb\u4e8e\u4efb\u4f55\u5e73\u53f0\u7684\u94bb\u5b54\u901a\u7528\u6570\u636e\u6a21\u578b\uff0c\u5c06\u5ca9\u6027\u3001\u6df1\u5ea6\u3001\u63cf\u8ff0\u7b49\u8981\u7d20\u62bd\u8c61\u4e3a\u5e73\u53f0\u65e0\u5173\u7684\u6570\u636e\u7ed3\u6784<\/span>\nUniversal Data Model:<\/strong> Designs a platform-independent universal borehole data model, abstracting lithology, depth, descriptions into platform-agnostic data structures<\/span><\/li>\n<li>\u5ca9\u6027\u82b1\u7eb9\u81ea\u52a8\u586b\u5145\u7b97\u6cd5\uff1a<\/strong>\u6839\u636e\u5ca9\u6027\u7f16\u7801\u81ea\u52a8\u5339\u914d\u6807\u51c6\u82b1\u7eb9\u5e93\uff0c\u6309\u6df1\u5ea6\u95f4\u9694\u81ea\u52a8\u586b\u5145\u5ca9\u6027\u67f1<\/span>\nAutomatic Lithology Pattern Fill:<\/strong> Automatically matches standard pattern libraries by lithology code and fills lithology columns by depth interval<\/span><\/li>\n<li>\u77e2\u91cf\u56fe\u5143\u751f\u6210\uff1a<\/strong>\u751f\u6210\u72ec\u7acb\u4e8e\u5e73\u53f0\u7684\u77e2\u91cf\u56fe\u5143\u6570\u636e\u96c6\uff0c\u53ef\u88ab\u4efb\u4f55\u652f\u6301\u77e2\u91cf\u7684\u8f6f\u4ef6\u6e32\u67d3<\/span>\nVector Primitive Generation:<\/strong> Generates platform-independent vector primitive datasets renderable by any vector-supporting software<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u8de8\u5e73\u53f0<\/span>Cross-Platform<\/span><\/span>\n\u81ea\u52a8\u6210\u56fe<\/span>Auto Plotting<\/span><\/span>\n\n<ul>\n<li>\u63d0\u51fa”\u4e00\u6b21\u751f\u6210\uff0c\u5230\u5904\u4f7f\u7528”\u7684\u8de8\u5e73\u53f0\u7406\u5ff5\uff0c\u7a81\u7834\u4e86CAD\u5e73\u53f0\u9501\u5b9a<\/span>\nProposes the “generate once, use everywhere” cross-platform concept, breaking through CAD platform lock-in<\/span><\/li>\n<li>\u5ca9\u6027\u82b1\u7eb9\u81ea\u52a8\u586b\u5145\u7b97\u6cd5\u89e3\u51b3\u4e86\u5927\u91cf\u94bb\u5b54\u6279\u91cf\u6210\u56fe\u7684\u5173\u952e\u6280\u672f\u74f6\u9888<\/span>\nThe automatic lithology pattern fill algorithm solves the key technical bottleneck for batch plotting of large numbers of boreholes<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box info\">\n\u6838\u5fc3\u4ef7\u503c\uff1a<\/span>Core Value:<\/span><\/strong>\n\u5c06\u6570\u636e\u4e0e\u8868\u73b0\u5206\u79bb\uff0c\u6570\u636e\u5b58\u50a8\u4e3a\u5e73\u53f0\u65e0\u5173\u7684\u4e2d\u95f4\u683c\u5f0f\uff0c\u56fe\u4ef6\u6e32\u67d3\u4ea4\u7531\u76ee\u6807\u5e73\u53f0\u5b8c\u6210\uff0c\u5b9e\u73b0\u4e86\u771f\u6b63\u7684\u8de8\u5e73\u53f0\u5e94\u7528\u3002<\/span>\nSeparates data from presentation: data is stored in a platform-independent intermediate format, and rendering is delegated to the target platform, achieving true cross-platform applicability.<\/span>\n<\/div>\n<\/div>\n\n<\/section>\n\n<section class=\"section\" id=\"conference\">\n\n<div class=\"section-icon\">\ud83d\udcdd<\/div>\n<h2>\u4f1a\u8bae\u8bba\u6587<\/span>Conference Papers<\/span><\/h2>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">Digitizing Drilling Logs – Challenges of typewritten forms<\/div>\n<div class=\"paper-meta\">\n\u4f1a\u8bae\uff1aGI\u4f1a\u8bae\u8bba\u6587\u96c6<\/span>\nConference: GI Conference Proceedings<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u5386\u53f2\u94bb\u5b54\u8bb0\u5f55\u5927\u91cf\u4f7f\u7528\u6253\u5b57\u673a\u586b\u5199\uff0c\u8fd9\u4e9b\u7eb8\u8d28\/\u626b\u63cf\u4ef6\u8868\u683c\u7684\u6570\u5b57\u5316\u9762\u4e34\u72ec\u7279\u6311\u6218\uff0c\u4e0e\u5370\u5237\u4f53\u6216\u624b\u5199\u4f53\u6587\u6863\u6709\u672c\u8d28\u4e0d\u540c\u3002<\/span>\nHistorical drilling logs were largely typewritten. Digitizing these paper\/scanned forms faces unique challenges fundamentally different from printed or handwritten documents.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5\u4e0e\u6280\u672f\u8def\u5f84<\/span>Core Methods & Technical Paths<\/span><\/h4>\n\u8bba\u6587\u5206\u6790\u4e86\u4e09\u5927\u6280\u672f\u8def\u5f84\uff1a<\/span>\nThe paper analyzes three technical paths:<\/span><\/p>\n\n<ol>\n<li>\u8868\u683c\u8bc6\u522b\u8def\u5f84<\/span>Table Recognition Path<\/span><\/strong><br\/>\n\u8bc6\u522b\u8868\u683c\u7ebf\u7ed3\u6784 \u2192 \u5b9a\u4f4d\u5355\u5143\u683c \u2192 \u63d0\u53d6\u5355\u5143\u683c\u5185\u6587\u5b57<\/span>\nRecognize table line structure \u2192 Locate cells \u2192 Extract text within cells<\/span><\/li>\n<li>\u8bed\u4e49\u5173\u8054\u8def\u5f84<\/span>Semantic Association Path<\/span><\/strong><br\/>\n\u8bc6\u522b\u5b57\u6bb5\u6807\u7b7e \u2192 \u5173\u8054\u90bb\u8fd1\u6570\u503c \u2192 \u6784\u5efa\u952e\u503c\u5bf9<\/span>\nRecognize field labels \u2192 Associate nearby values \u2192 Build key-value pairs<\/span><\/li>\n<li>\u6570\u503c\u63d0\u53d6\u8def\u5f84<\/span>Numerical Extraction Path<\/span><\/strong><br\/>\n\u9488\u5bf9\u6df1\u5ea6\u3001\u539a\u5ea6\u3001\u767e\u5206\u6bd4\u7b49\u6570\u503c\u578b\u5b57\u6bb5\u7684\u7279\u6b8a\u63d0\u53d6\u7b56\u7565<\/span>\nSpecial extraction strategies for numerical fields like depth, thickness, percentages<\/span><\/li>\n<\/ol>\n\n<h4>\u5173\u952e\u6311\u6218<\/span>Key Challenges<\/span><\/h4>\n<ul>\n<li>\u6253\u5b57\u673a\u5b57\u4f53\u7b49\u5bbd\u7279\u6027\u5bfc\u81f4\u7684\u5217\u5bf9\u9f50\u6b67\u4e49<\/span>\nColumn alignment ambiguity caused by monospace typewriter fonts<\/span><\/li>\n<li>\u590d\u5199\u7eb8\/\u78b3\u5e26\u9020\u6210\u7684\u5b57\u7b26\u7c98\u8fde\u548c\u6a21\u7cca<\/span>\nCharacter adhesion and blurring from carbon paper\/ribbon<\/span><\/li>\n<li>\u8868\u683c\u7ebf\u7f3a\u5931\u6216\u65ad\u88c2\u5bfc\u81f4\u7684\u7ed3\u6784\u8bc6\u522b\u56f0\u96be<\/span>\nStructural recognition difficulty due to missing or broken table lines<\/span><\/li>\n<li>\u8bed\u4e49\u5173\u8054\u9519\u8bef\uff08\u5982\u6df1\u5ea6\u503c\u4e0e\u5ca9\u6027\u63cf\u8ff0\u9519\u914d\uff09<\/span>\nSemantic association errors (e.g., depth values mismatched with lithology descriptions)<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box warn\">\n\u5b66\u672f\u4ef7\u503c\uff1a<\/span>Academic Value:<\/span><\/strong>\n\u8be5\u6587\u662f\u5c11\u6570\u4e13\u95e8\u8ba8\u8bba\u6253\u5b57\u673a\u586b\u5199\u8868\u683c\u6570\u5b57\u5316\u6311\u6218\u7684\u6587\u732e\uff0c\u5bf9\u5904\u740620\u4e16\u7eaa\u5386\u53f2\u94bb\u5b54\u6863\u6848\u5177\u6709\u91cd\u8981\u53c2\u8003\u4ef7\u503c\u3002<\/span>\nThis is one of few papers specifically addressing typewritten form digitization challenges, providing important reference for processing 20th-century historical drilling archives.<\/span>\n<\/div>\n<\/div>\n\n<\/section>\n\n<section class=\"section\" id=\"comparison\">\n\n<div class=\"section-icon\">\u2696\ufe0f<\/div>\n<h2>\u65b9\u6cd5\u5bf9\u6bd4\u4e0e\u521b\u65b0\u603b\u7ed3<\/span>Method Comparison & Innovation Summary<\/span><\/h2>\n\n<h3>\u8bba\u6587\u65b9\u6cd5\u7279\u5f81\u5bf9\u6bd4<\/span>Paper Method Characteristics Comparison<\/span><\/h3>\n\n<div class=\"table-wrap\">\n<table>\n<thead>\n<tr>\n<th>\u8bba\u6587<\/span>Paper<\/span><\/th>\n<th>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Method<\/span><\/th>\n<th>\u6280\u672f\u8def\u7ebf<\/span>Technical Route<\/span><\/th>\n<th>\u4e3b\u8981\u521b\u65b0<\/span>Main Innovation<\/span><\/th>\n<th>\u8bc4\u4ef7\u6307\u6807<\/span>Evaluation Metric<\/span><\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>dh2loop 1.0<\/td>\n<td>\u8bcd\u5e93+\u6a21\u7cca\u5339\u914d<\/span>Lexicon + Fuzzy Matching<\/span><\/td>\n<td>NLP+\u89c4\u5219<\/span>NLP + Rules<\/span><\/td>\n<td>\u591a\u7ea7\u5ca9\u6027\u5206\u7c7b\u5f00\u6e90\u5e93<\/span>Multi-level lithology classification open-source library<\/span><\/td>\n<td>\u63d0\u53d6\u738793%\/86%\/34%<\/span>Extraction rate 93%\/86%\/34%<\/span><\/td>\n<\/tr>\n<tr>\n<td>litho2strat 1.0<\/td>\n<td>\u641c\u7d22\u7b97\u6cd5+\u8de8\u5b54\u5bf9\u6bd4<\/span>Search + Cross-hole Correlation<\/span><\/td>\n<td>\u7ea6\u675f\u641c\u7d22+\u62d3\u6251\u5206\u6790<\/span>Constrained Search + Topology<\/span><\/td>\n<td>\u5730\u5c42\u89e3\u91ca\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Stratigraphic interpretation uncertainty quantification<\/span><\/td>\n<td>\u5730\u8d28\u4e00\u81f4\u6027\u6982\u7387<\/span>Geological consistency probability<\/span><\/td>\n<\/tr>\n<tr>\n<td>\u591a\u5e03\u5c40\u6295\u7968OCR<\/span>Multi-layout Voting OCR<\/span><\/td>\n<td>\u591a\u5e03\u5c40\u8c03\u6574+\u6295\u7968<\/span>Multi-layout + Voting<\/span><\/td>\n<td>\u56fe\u50cf\u5904\u7406+\u96c6\u6210\u5b66\u4e60<\/span>Image Processing + Ensemble<\/span><\/td>\n<td>\u590d\u6742\u5e03\u5c40OCR\u4f18\u5316\u8303\u5f0f<\/span>Complex layout OCR optimization paradigm<\/span><\/td>\n<td>F1=97.96%\/94.36%<\/span>F1=97.96%\/94.36%<\/span><\/td>\n<\/tr>\n<tr>\n<td>\u6c99\u6d41\u6cb3\u5e45\u6570\u636e\u5e93<\/span>Shaliuhe Database<\/span><\/td>\n<td>\u5206\u5c42\u6570\u636e\u6a21\u578b+\u8d28\u63a7<\/span>Hierarchical Model + QC<\/span><\/td>\n<td>\u6570\u636e\u5e93\u5de5\u7a0b<\/span>Database Engineering<\/span><\/td>\n<td>\u533a\u57df\u94bb\u5b54\u6570\u636e\u5e93\u4f53\u7cfb<\/span>Regional borehole database system<\/span><\/td>\n<td>\u6570\u636e\u5b8c\u6574\u6027<\/span>Data completeness<\/span><\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5e73\u53f0\u81ea\u52a8\u751f\u6210<\/span>Cross-platform Auto Generation<\/span><\/td>\n<td>\u901a\u7528\u6a21\u578b+\u77e2\u91cf\u751f\u6210<\/span>Universal Model + Vector Generation<\/span><\/td>\n<td>\u6570\u636e\u4e2d\u95f4\u4ef6<\/span>Data Middleware<\/span><\/td>\n<td>\u5e73\u53f0\u65e0\u5173\u7684\u6210\u56fe\u65b9\u6cd5<\/span>Platform-independent plotting<\/span><\/td>\n<td>\u8de8\u5e73\u53f0\u517c\u5bb9\u6027<\/span>Cross-platform compatibility<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n\n<h3>\u6280\u672f\u53d1\u5c55\u8d8b\u52bf<\/span>Technology Development Trends<\/span><\/h3>\n\n<div class=\"tip-box\">\n\u8d8b\u52bf\u4e00\uff1a\u4ece\u89c4\u5219\u9a71\u52a8\u5230\u6570\u636e\u9a71\u52a8<\/span>Trend 1: From Rule-driven to Data-driven<\/span><\/strong><br\/>\n\u65e9\u671f\u7814\u7a76\u4f9d\u8d56\u6b63\u5219\u8868\u8fbe\u5f0f\u548c\u89c4\u5219\u5339\u914d\uff08\u5982dh2loop\u7684\u8bcd\u5e93\u65b9\u6cd5\uff09\uff0c\u6700\u65b0\u7814\u7a76\u5f00\u59cb\u5f15\u5165XGBoost\u3001\u89c6\u89c9\u5927\u6a21\u578b\u7b49\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u3002<\/span>\nEarly research relied on regex and rule matching (e.g., dh2loop’s lexicon method); latest research introduces XGBoost, vision LLMs, and other ML approaches.<\/span>\n<\/div>\n\n<div class=\"tip-box\">\n\u8d8b\u52bf\u4e8c\uff1a\u4ece\u786e\u5b9a\u6027\u5230\u6982\u7387\u5316<\/span>Trend 2: From Deterministic to Probabilistic<\/span><\/strong><br\/>\nlitho2strat\u5c06\u5730\u5c42\u89e3\u91ca\u7684\u4e0d\u786e\u5b9a\u6027\u663e\u5f0f\u91cf\u5316\uff0c\u4ee3\u8868\u4e86\u4ece\u5355\u4e00\u6700\u4f18\u89e3\u5230\u6982\u7387\u89e3\u96c6\u7684\u65b9\u6cd5\u8bba\u8f6c\u53d8\u3002<\/span>\nlitho2strat explicitly quantifies stratigraphic interpretation uncertainty, representing a methodological shift from single optimal solutions to probabilistic solution sets.<\/span>\n<\/div>\n\n<div class=\"tip-box\">\n\u8d8b\u52bf\u4e09\uff1a\u4ece\u5355\u70b9\u5de5\u5177\u5230\u5f00\u6e90\u751f\u6001<\/span>Trend 3: From Point Tools to Open Ecosystems<\/span><\/strong><br\/>\ndh2loop\u3001litho2strat\u7b49\u53d1\u8868\u4e8eGMD\u7684\u8bba\u6587\u5747\u63d0\u4f9b\u5b8c\u6574\u5f00\u6e90\u5b9e\u73b0\uff0c\u5b66\u672f\u6210\u679c\u5411\u5de5\u7a0b\u5de5\u5177\u8f6c\u5316\u7684\u5468\u671f\u5927\u5e45\u7f29\u77ed\u3002<\/span>\nPapers published in GMD (dh2loop, litho2strat) provide complete open-source implementations, significantly shortening the cycle from academic results to engineering tools.<\/span>\n<\/div>\n\n<\/section>\n\n<div class=\"footnotes\">\n \u94bb\u5b54\u67f1\u72b6\u56fe\u8bba\u6587\u65b9\u6cd5\u5206\u6790 | \u57fa\u4e8e\u516c\u5f00\u5b66\u672f\u6587\u732e\u6574\u7406<\/span>Borehole Log Paper Analysis | Compiled from public academic literature<\/span><\/p>\n<\/div>\n\n<\/div>\n\n\n<script>\nfunction switchLanguage(lang) {\n document.documentElement.lang = lang;\n document.body.className = (lang === 'en') ? 'en-active' : '';\n \n var btnZh = document.getElementById('btn-zh');\n var btnEn = document.getElementById('btn-en');\n \n if (lang === 'zh') {\n btnZh.classList.add('active');\n btnEn.classList.remove('active');\n } else {\n btnEn.classList.add('active');\n btnZh.classList.remove('active');\n }\n}\n\ndocument.addEventListener('DOMContentLoaded', function() {\n switchLanguage('zh');\n});\n\ndocument.querySelectorAll('a[href^=\"#\"]').forEach(function(anchor) {\n anchor.addEventListener('click', function(e) {\n e.preventDefault();\n var target = document.querySelector(this.getAttribute('href'));\n if (target) {\n target.scrollIntoView({ behavior: 'smooth', block: 'start' });\n }\n });\n});\n<\/script>\n\n<style media=\"print\">\n .lang-bar { display: none !important; }\n section { page-break-inside: avoid; }\n<\/style>\n\n<\/body>\n<\/html>\n\n","protected":false},"excerpt":{"rendered":"\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0Review of Borehole Log Digitization Papers […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11,9],"tags":[],"class_list":["post-290","post","type-post","status-publish","format-standard","hentry","category-cad","category-ai"],"blocksy_meta":[],"_links":{"self":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts\/290","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=290"}],"version-history":[{"count":1,"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts\/290\/revisions"}],"predecessor-version":[{"id":291,"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts\/290\/revisions\/291"}],"wp:attachment":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=290"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=290"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=290"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}

{"id":290,"date":"2026-07-02T16:17:34","date_gmt":"2026-07-02T08:17:34","guid":{"rendered":"https:\/\/numsimlab.com\/?p=290"},"modified":"2026-07-02T16:17:34","modified_gmt":"2026-07-02T08:17:34","slug":"%e9%92%bb%e5%ad%94%e6%9f%b1%e7%8a%b6%e5%9b%be%e6%95%b0%e5%ad%97%e5%8c%96%e8%ae%ba%e6%96%87%e7%bb%bc%e8%bf%b0","status":"publish","type":"post","link":"https:\/\/numsimlab.com\/?p=290","title":{"rendered":"\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0"},"content":{"rendered":"\n\n\n\n\n\n\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0<\/span>Review of Borehole Log Digitization Papers<\/span><\/title>\n<style>\n:root {\n --bg: #f5faf7;\n --bg2: #eef6f1;\n --ink: #1a2e23;\n --muted: #5a7a6e;\n --rule: #c0d8cc;\n --accent: #059669;\n --accent2: #0d9488;\n --code-bg: #1e2d26;\n --code-fg: #d1fae5;\n}\n\n* { box-sizing: border-box; margin: 0; padding: 0; }\n\nhtml { scroll-behavior: smooth; }\n\nbody {\n font-family: \"Microsoft YaHei\", \"PingFang SC\", system-ui, -apple-system, sans-serif;\n background: var(--bg);\n color: var(--ink);\n line-height: 1.8;\n}\n\nbody.en-active { font-family: system-ui, -apple-system, sans-serif; }\n\n\/* Language Switcher Bar *\/\n.lang-bar {\n position: sticky;\n top: 0;\n z-index: 1000;\n background: var(--accent);\n display: flex;\n justify-content: center;\n gap: 1rem;\n padding: 0.6rem 1rem;\n box-shadow: 0 2px 8px rgba(0,0,0,0.15);\n}\n\n.lang-bar button {\n background: rgba(255,255,255,0.2);\n border: 2px solid transparent;\n color: #fff;\n padding: 0.4rem 1.2rem;\n border-radius: 4px;\n cursor: pointer;\n font-size: 0.95rem;\n font-weight: 600;\n transition: all 0.2s;\n}\n\n.lang-bar button:hover {\n background: rgba(255,255,255,0.35);\n}\n\n.lang-bar button.active {\n background: #fff;\n color: var(--accent);\n border-color: #fff;\n}\n\n\/* Container *\/\n.container {\n max-width: 1100px;\n margin: 0 auto;\n padding: 0 1.5rem;\n}\n\n\/* Cover *\/\n.cover {\n text-align: center;\n padding: 4rem 1rem 3rem;\n background: var(--bg2);\n border-bottom: 1px solid var(--rule);\n}\n\n.cover h1 {\n font-size: 2.2rem;\n color: var(--ink);\n margin-bottom: 0.8rem;\n font-weight: 700;\n}\n\n.cover .subtitle {\n font-size: 1.3rem;\n color: var(--accent);\n font-weight: 600;\n margin-bottom: 0.6rem;\n}\n\n.cover .desc {\n font-size: 1rem;\n color: var(--muted);\n}\n\n\/* TOC *\/\n.toc {\n background: #fff;\n border: 1px solid var(--rule);\n border-radius: 8px;\n padding: 1.5rem 2rem;\n margin: 2rem 0;\n}\n\n.toc h2 {\n font-size: 1.3rem;\n margin-bottom: 1rem;\n color: var(--accent);\n}\n\n.toc ol, .toc ul {\n padding-left: 1.5rem;\n}\n\n.toc li {\n margin: 0.4rem 0;\n}\n\n.toc a {\n color: var(--ink);\n text-decoration: none;\n transition: color 0.2s;\n}\n\n.toc a:hover {\n color: var(--accent);\n text-decoration: underline;\n}\n\n\/* Sections *\/\nsection {\n background: #fff;\n border: 1px solid var(--rule);\n border-radius: 8px;\n padding: 2rem 2.5rem;\n margin: 2rem 0;\n}\n\nsection h2 {\n font-size: 1.6rem;\n color: var(--accent);\n margin-bottom: 1.2rem;\n padding-bottom: 0.5rem;\n border-bottom: 2px solid var(--rule);\n}\n\nsection h3 {\n font-size: 1.25rem;\n color: var(--accent2);\n margin: 1.5rem 0 0.8rem;\n}\n\nsection h4 {\n font-size: 1.1rem;\n color: var(--ink);\n margin: 1.2rem 0 0.6rem;\n font-weight: 600;\n}\n\nsection p {\n margin-bottom: 0.9rem;\n text-align: justify;\n}\n\n\/* Tables *\/\ntable {\n width: 100%;\n border-collapse: collapse;\n margin: 1.2rem 0;\n font-size: 0.95rem;\n}\n\nth, td {\n border: 1px solid var(--rule);\n padding: 0.7rem 0.9rem;\n text-align: left;\n}\n\nth {\n background: var(--bg2);\n font-weight: 600;\n color: var(--ink);\n}\n\ntr:nth-child(even) { background: var(--bg); }\n\n\/* Code *\/\npre {\n background: var(--code-bg);\n color: var(--code-fg);\n padding: 1.2rem 1.5rem;\n border-radius: 6px;\n overflow-x: auto;\n font-family: Consolas, \"Courier New\", monospace;\n font-size: 0.9rem;\n line-height: 1.6;\n margin: 1rem 0;\n}\n\ncode {\n font-family: Consolas, \"Courier New\", monospace;\n background: rgba(5,150,105,0.1);\n padding: 0.15rem 0.4rem;\n border-radius: 3px;\n font-size: 0.9em;\n color: var(--accent);\n}\n\npre code {\n background: transparent;\n padding: 0;\n color: inherit;\n}\n\n\/* Tip box *\/\n.tip-box {\n background: #ecfdf5;\n border-left: 4px solid var(--accent);\n padding: 1rem 1.2rem;\n margin: 1.2rem 0;\n border-radius: 0 6px 6px 0;\n font-size: 0.95rem;\n}\n\n.tip-box.warn {\n background: #fffbeb;\n border-left-color: #f59e0b;\n}\n\n.tip-box.info {\n background: #eff6ff;\n border-left-color: #3b82f6;\n}\n\n.tip-box strong { color: var(--accent); }\n.tip-box.warn strong { color: #b45309; }\n.tip-box.info strong { color: #1d4ed8; }\n\n\/* Highlight *\/\nmark.key {\n background: #d1fae5;\n color: #065f46;\n padding: 0.15em 0.4em;\n border-radius: 3px;\n font-weight: 700;\n}\n\n\/* Lists *\/\nul, ol {\n padding-left: 1.8rem;\n margin: 0.8rem 0;\n}\n\nli { margin: 0.4rem 0; }\n\n\/* Footnotes *\/\n.footnotes {\n font-size: 0.85rem;\n color: var(--muted);\n border-top: 1px solid var(--rule);\n margin-top: 2rem;\n padding-top: 1rem;\n text-align: center;\n}\n\n\/* Table wrap for scrolling *\/\n.table-wrap {\n overflow-x: auto;\n margin-bottom: 1rem;\n}\n\n\/* Section icon *\/\n.section-icon {\n width: 32px;\n height: 32px;\n border-radius: 8px;\n background: var(--bg2);\n display: inline-flex;\n align-items: center;\n justify-content: center;\n font-size: 1rem;\n margin-right: 0.6rem;\n vertical-align: middle;\n}\n\n\/* Paper card *\/\n.paper-card {\n background: var(--bg2);\n border: 1px solid var(--rule);\n border-radius: 8px;\n padding: 1.5rem;\n margin: 1.2rem 0;\n}\n\n.paper-title {\n font-size: 1.15rem;\n font-weight: 700;\n color: var(--ink);\n margin-bottom: 0.5rem;\n}\n\n.paper-meta {\n font-size: 0.9rem;\n color: var(--muted);\n margin-bottom: 1rem;\n}\n\n\/* Method flow *\/\n.method-flow {\n display: flex;\n align-items: center;\n flex-wrap: wrap;\n gap: 0.5rem;\n margin: 1rem 0;\n padding: 1rem;\n background: var(--bg);\n border-radius: 6px;\n}\n\n.method-step {\n background: #fff;\n border: 1px solid var(--rule);\n padding: 0.4rem 0.8rem;\n border-radius: 4px;\n font-size: 0.9rem;\n font-weight: 600;\n}\n\n.method-arrow {\n color: var(--accent);\n font-weight: bold;\n}\n\n\/* Innovation tags *\/\n.innovation-tag {\n display: inline-block;\n background: #d1fae5;\n color: #065f46;\n padding: 0.2em 0.6em;\n border-radius: 4px;\n font-size: 0.8rem;\n font-weight: 600;\n margin-right: 0.4rem;\n margin-bottom: 0.3rem;\n}\n\n\/* Responsive *\/\n@media (max-width: 768px) {\n .cover h1 { font-size: 1.5rem; }\n .cover .subtitle { font-size: 1.05rem; }\n section { padding: 1.2rem 1rem; }\n section h2 { font-size: 1.3rem; }\n table { font-size: 0.85rem; }\n th, td { padding: 0.5rem; }\n .method-flow { flex-direction: column; align-items: flex-start; }\n}\n\n\/* Language display control *\/\n[data-lang=\"en\"] { display: none; }\nhtml[lang=\"en\"] [data-lang=\"zh\"] { display: none; }\nhtml[lang=\"en\"] [data-lang=\"en\"] { display: inline; }\n\nh2 [data-lang=\"en\"], h3 [data-lang=\"en\"], h4 [data-lang=\"en\"], p [data-lang=\"en\"], li [data-lang=\"en\"], th [data-lang=\"en\"], td [data-lang=\"en\"], div [data-lang=\"en\"], strong [data-lang=\"en\"] {\n display: none;\n}\n\nhtml[lang=\"en\"] h2 [data-lang=\"zh\"], html[lang=\"en\"] h3 [data-lang=\"zh\"], html[lang=\"en\"] h4 [data-lang=\"zh\"], html[lang=\"en\"] p [data-lang=\"zh\"], html[lang=\"en\"] li [data-lang=\"zh\"], html[lang=\"en\"] th [data-lang=\"zh\"], html[lang=\"en\"] td [data-lang=\"zh\"], html[lang=\"en\"] div [data-lang=\"zh\"], html[lang=\"en\"] strong [data-lang=\"zh\"] {\n display: none;\n}\n\nhtml[lang=\"en\"] h2 [data-lang=\"en\"], html[lang=\"en\"] h3 [data-lang=\"en\"], html[lang=\"en\"] h4 [data-lang=\"en\"], html[lang=\"en\"] p [data-lang=\"en\"], html[lang=\"en\"] li [data-lang=\"en\"], html[lang=\"en\"] th [data-lang=\"en\"], html[lang=\"en\"] td [data-lang=\"en\"], html[lang=\"en\"] div [data-lang=\"en\"], html[lang=\"en\"] strong [data-lang=\"en\"] {\n display: inline;\n}\n\n.toc h2 span, .cover h1 span, .cover .subtitle span, .cover .desc span, section h2 span, section h3 span, section h4 span {\n display: inline;\n}\n\nhtml[lang=\"en\"] .toc h2 [data-lang=\"zh\"], html[lang=\"en\"] .cover h1 [data-lang=\"zh\"], html[lang=\"en\"] .cover .subtitle [data-lang=\"zh\"], html[lang=\"en\"] .cover .desc [data-lang=\"zh\"], html[lang=\"en\"] section h2 [data-lang=\"zh\"], html[lang=\"en\"] section h3 [data-lang=\"zh\"], html[lang=\"en\"] section h4 [data-lang=\"zh\"] {\n display: none;\n}\n\nhtml[lang=\"en\"] .toc h2 [data-lang=\"en\"], html[lang=\"en\"] .cover h1 [data-lang=\"en\"], html[lang=\"en\"] .cover .subtitle [data-lang=\"en\"], html[lang=\"en\"] .cover .desc [data-lang=\"en\"], html[lang=\"en\"] section h2 [data-lang=\"en\"], html[lang=\"en\"] section h3 [data-lang=\"en\"], html[lang=\"en\"] section h4 [data-lang=\"en\"] {\n display: inline;\n}\n<\/style>\n<\/head>\n<body>\n\n\n<div class=\"lang-bar\">\n <button id=\"btn-zh\" class=\"active\" onclick=\"switchLanguage('zh')\">\n \u4e2d\u6587<\/span>Chinese<\/span>\n <\/button>\n <button id=\"btn-en\" onclick=\"switchLanguage('en')\">\n English<\/span>English<\/span>\n <\/button>\n<\/div>\n\n<div class=\"container\">\n\n\n<div class=\"cover\">\n <h1>\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0<\/span>Review of Borehole Log Digitization Papers<\/span><\/h1>\n <div class=\"subtitle\">Academic Research Review<\/div>\n <div class=\"desc\">\u7cfb\u7edf\u68b3\u7406\u94bb\u5b54\u6570\u636e\u6570\u5b57\u5316\u3001\u5730\u5c42\u89e3\u91ca\u3001OCR\u8bc6\u522b\u7b49\u9886\u57df\u6838\u5fc3\u8bba\u6587\u7684\u65b9\u6cd5\u8bba\u4e0e\u521b\u65b0\u8d21\u732e\u3002<\/span>Systematic review of methodologies and innovations in borehole data digitization, stratigraphic interpretation, and OCR recognition.<\/span><\/div>\n<\/div>\n\n\n<nav class=\"toc\">\n <h2>\u76ee\u5f55<\/span>Table of Contents<\/span><\/h2>\n <ol>\n <li><a href=\"#overview\">\u7814\u7a76\u8109\u7edc\u6982\u8ff0<\/span>Research Overview<\/span><\/a><\/li>\n <li><a href=\"#international\">\u56fd\u9645\u671f\u520a\u8bba\u6587<\/span>International Journal Papers<\/span><\/a><\/li>\n <li><a href=\"#chinese\">\u4e2d\u6587\u6838\u5fc3\u7814\u7a76<\/span>Chinese Core Research<\/span><\/a><\/li>\n <li><a href=\"#conference\">\u4f1a\u8bae\u8bba\u6587<\/span>Conference Papers<\/span><\/a><\/li>\n <li><a href=\"#comparison\">\u65b9\u6cd5\u5bf9\u6bd4\u4e0e\u521b\u65b0\u603b\u7ed3<\/span>Method Comparison & Innovation Summary<\/span><\/a><\/li>\n <\/ol>\n<\/nav>\n\n<section class=\"section\" id=\"overview\">\n\n<div class=\"section-icon\">\ud83d\udcd6<\/div>\n<h2>\u7814\u7a76\u8109\u7edc\u6982\u8ff0<\/span>Research Overview<\/span><\/h2>\n\n\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u76f8\u5173\u7684\u5b66\u672f\u7814\u7a76\u53ef\u5206\u4e3a\u56db\u5927\u4e3b\u7ebf\uff1a<\/span>\nAcademic research on borehole log digitization can be divided into four main threads:<\/span><\/p>\n\n<ol>\n<li>\u81ea\u52a8\u5316\u6570\u636e\u63d0\u53d6<\/span>Automated Data Extraction<\/span><\/strong> \u2014 \n\u4ece\u975e\u7ed3\u6784\u5316\u5730\u8d28\u62a5\u544a(PDF\/\u626b\u63cf\u4ef6)\u4e2d\u81ea\u52a8\u63d0\u53d6\u5b54\u53e3\u3001\u6d4b\u659c\u3001\u5ca9\u6027\u7b49\u7ed3\u6784\u5316\u6570\u636e<\/span>\nAutomatically extracting structured data (collar, survey, lithology) from unstructured geological reports (PDF\/scans)<\/span><\/li>\n<li>\u5730\u5c42\u5e8f\u5217\u6062\u590d\u4e0e\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Stratigraphic Sequence Recovery & Uncertainty Quantification<\/span><\/strong> \u2014 \n\u4ece\u5ca9\u6027\u63cf\u8ff0\u4e2d\u6062\u590d\u5730\u5c42\u5e8f\u5217\uff0c\u8bc4\u4f30\u89e3\u91ca\u7ed3\u679c\u7684\u4e0d\u786e\u5b9a\u6027<\/span>\nRecovering stratigraphic sequences from lithological descriptions and quantifying interpretation uncertainty<\/span><\/li>\n<li>OCR\u4e0e\u56fe\u50cf\u8bc6\u522b\u4f18\u5316<\/span>OCR & Image Recognition Optimization<\/span><\/strong> \u2014 \n\u63d0\u5347\u94bb\u5b54\u67f1\u72b6\u56fe\u56fe\u50cf\u4e2d\u6587\u5b57\u548c\u7b26\u53f7\u7684\u8bc6\u522b\u51c6\u786e\u7387<\/span>\nImproving recognition accuracy of text and symbols in borehole log images<\/span><\/li>\n<li>\u6570\u636e\u5e93\u5efa\u8bbe\u4e0e\u8de8\u5e73\u53f0\u5e94\u7528<\/span>Database Construction & Cross-Platform Applications<\/span><\/strong> \u2014 \n\u5efa\u7acb\u6807\u51c6\u5316\u7684\u94bb\u5b54\u6570\u636e\u5e93\uff0c\u5b9e\u73b0\u6570\u636e\u5171\u4eab\u548c\u8de8\u5e73\u53f0\u53ef\u89c6\u5316<\/span>\nEstablishing standardized borehole databases for data sharing and cross-platform visualization<\/span><\/li>\n<\/ol>\n\n<\/section>\n\n<section class=\"section\" id=\"international\">\n\n<div class=\"section-icon\">\ud83c\udf0d<\/div>\n<h2>\u56fd\u9645\u671f\u520a\u8bba\u6587<\/span>International Journal Papers<\/span><\/h2>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">dh2loop 1.0: an open-source python library for automated processing and classification of geological logs<\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1aGeoscientific Model Development (2026) | \u4f5c\u8005\uff1aRanee Joshi \u7b49<\/span>\nJournal: Geoscientific Model Development (2026) | Authors: Ranee Joshi et al.<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u5730\u8d28\u94bb\u5b54\u62a5\u544a\u4ee5\u975e\u7ed3\u6784\u5316\u6587\u672c\u5f62\u5f0f\u5b58\u50a8\u4e8e\u5404\u5730\u8d28\u8c03\u67e5\u5c40\u6570\u636e\u5e93\u4e2d\uff0c\u5ca9\u6027\u63cf\u8ff0\u5b58\u5728\u4e25\u91cd\u7684\u547d\u540d\u4e0d\u4e00\u81f4\u95ee\u9898\uff08\u540c\u4e00\u5ca9\u6027\u53ef\u80fd\u6709\u6570\u5341\u79cd\u4e0d\u540c\u63cf\u8ff0\u65b9\u5f0f\uff09\uff0c\u963b\u788d\u4e86\u5927\u89c4\u6a21\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u7684\u6570\u636e\u51c6\u5907\u3002<\/span>\nGeological drillhole reports are stored as unstructured text in geological survey databases. Lithological descriptions suffer from severe naming inconsistency (the same lithology may have dozens of different descriptions), hindering data preparation for large-scale 3D geological modeling.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<div class=\"method-flow\">\nPDF\u89e3\u6790<\/span>PDF Parsing<\/span><\/span>\n\u2192<\/span>\n\u6587\u672c\u5206\u5757<\/span>Text Chunking<\/span><\/span>\n\u2192<\/span>\n\u6b63\u5219\u5339\u914d<\/span>Regex Matching<\/span><\/span>\n\u2192<\/span>\n\u8bcd\u5e93\u5339\u914d<\/span>Lexicon Matching<\/span><\/span>\n\u2192<\/span>\n\u6a21\u7cca\u5b57\u7b26\u4e32\u5339\u914d<\/span>Fuzzy Matching<\/span><\/span>\n\u2192<\/span>\n\u591a\u7ea7\u5206\u7c7b<\/span>Multi-level Classification<\/span><\/span>\n<\/div>\n\n<ul>\n<li>\u4e09\u6bb5\u5f0f\u6570\u636e\u6a21\u578b\uff1a<\/strong>\u5c06\u94bb\u5b54\u6570\u636e\u6807\u51c6\u5316\u4e3a\u5b54\u53e3(collar)\u3001\u6d4b\u659c(survey)\u3001\u5ca9\u6027(lithology)\u4e09\u4e2a\u6838\u5fc3\u95f4\u9694\u8868<\/span>\nThree-part Data Model:<\/strong> Standardizes drillhole data into collar, survey, and lithology interval tables<\/span><\/li>\n<li>\u8bcd\u5e93\u9a71\u52a8\u5339\u914d\uff1a<\/strong>\u6784\u5efa\u6807\u51c6\u5316\u5ca9\u6027\u8bcd\u5e93\uff0c\u5c06\u539f\u59cb\u63cf\u8ff0\u6620\u5c04\u5230\u6807\u51c6\u672f\u8bed<\/span>\nLexicon-driven Matching:<\/strong> Builds standardized lithology lexicon to map raw descriptions to standard terms<\/span><\/li>\n<li>\u6a21\u7cca\u5339\u914d\u7b97\u6cd5\uff1a<\/strong>\u91c7\u7528\u6a21\u7cca\u5b57\u7b26\u4e32\u5339\u914d(fuzzy string matching)\u5904\u7406\u62fc\u5199\u53d8\u4f53\u548c\u540c\u4e49\u8bcd<\/span>\nFuzzy Matching:<\/strong> Uses fuzzy string matching to handle spelling variants and synonyms<\/span><\/li>\n<li>\u591a\u7ea7\u5206\u7c7b\u4f53\u7cfb\uff1a<\/strong>\u652f\u6301\u4ece\u8be6\u7ec6\u5ca9\u6027\u5230\u7c97\u7c92\u5ea6\u7c7b\u522b\u7684\u591a\u7ea7\u5f52\u5e76\uff0c\u9002\u914d\u4e0d\u540c\u5c3a\u5ea6\u7684\u4e09\u7ef4\u5efa\u6a21\u9700\u6c42<\/span>\nMulti-level Taxonomy:<\/strong> Supports aggregation from detailed lithology to coarse categories for different modeling scales<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u5f00\u6e90\u5b9e\u73b0<\/span>Open Source<\/span><\/span>\n\u8bcd\u5e93\u6807\u51c6\u5316<\/span>Lexicon Standardization<\/span><\/span>\n\u591a\u7ea7\u5206\u7c7b<\/span>Hierarchical Classification<\/span><\/span>\n\n<ul>\n<li>\u9996\u6b21\u5c06\u5b8c\u6574\u7684\u94bb\u5b54\u6570\u636e\u63d0\u53d6\u6d41\u7a0b\u5f00\u6e90\u4e3a\u53ef\u590d\u7528Python\u5e93<\/span>\nFirst open-source Python library providing a complete drillhole data extraction pipeline<\/span><\/li>\n<li>\u63d0\u51fa\u8bcd\u5e93+\u6a21\u7cca\u5339\u914d\u7684\u4e24\u9636\u6bb5\u5ca9\u6027\u547d\u540d\u6d88\u6b67\u65b9\u6cd5<\/span>\nProposes a two-stage lithology name disambiguation method: lexicon + fuzzy matching<\/span><\/li>\n<li>\u591a\u7ea7\u5ca9\u6027\u5206\u7c7b\u673a\u5236\u89e3\u51b3\u4e86\u4ece\u533a\u57df\u5efa\u6a21\u5230\u5c40\u90e8\u5efa\u6a21\u7684\u6570\u636e\u7c92\u5ea6\u9002\u914d\u95ee\u9898<\/span>\nMulti-level classification mechanism solves data granularity adaptation from regional to local modeling<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box\">\n\u5b9e\u9a8c\u7ed3\u679c\uff1a<\/span>Experimental Results:<\/span><\/strong><br\/>\n\u5b54\u53e3\u6570\u636e\u63d0\u53d6\u7387 93%<\/mark>\uff0c\u6d4b\u659c\u6570\u636e 86%<\/mark>\uff0c\u5ca9\u6027\u6570\u636e 34%<\/mark>\uff1b\u5176\u4e2d86%\u7684\u5ca9\u6027\u6570\u636e\u53ef\u6210\u529f\u5339\u914d\u5230\u6807\u51c6\u8bcd\u5e93\u3002<\/span>\nCollar extraction rate 93%<\/mark>, survey 86%<\/mark>, lithology 34%<\/mark>; 86% of lithology data successfully matched to standard lexicon.<\/span>\n<\/div>\n<\/div>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">Automated stratigraphic interpretation from drillhole lithological descriptions with uncertainty quantification: litho2strat 1.0<\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1aGeoscientific Model Development (2026)<\/span>\nJournal: Geoscientific Model Development (2026)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u4ece\u94bb\u5b54\u5ca9\u6027\u6570\u636e\u6062\u590d\u5730\u5c42\u5e8f\u5217\u662f\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u7684\u5173\u952e\u6b65\u9aa4\uff0c\u4f46\u4f20\u7edf\u4eba\u5de5\u89e3\u91ca\u6548\u7387\u4f4e\u4e14\u4e3b\u89c2\u6027\u5f3a\u3002\u5982\u4f55\u81ea\u52a8\u4ece\u5ca9\u6027\u63cf\u8ff0\u4e2d\u63a8\u65ad\u5730\u5c42\u5e8f\u5217\uff0c\u5e76\u91cf\u5316\u89e3\u91ca\u7684\u4e0d\u786e\u5b9a\u6027\uff0c\u662f\u6838\u5fc3\u79d1\u5b66\u95ee\u9898\u3002<\/span>\nRecovering stratigraphic sequences from drillhole lithology data is a key step in 3D geological modeling, but traditional manual interpretation is inefficient and subjective. Automatically inferring stratigraphic sequences from lithological descriptions while quantifying uncertainty is the core scientific problem.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<div class=\"method-flow\">\n\u5ca9\u6027\u8f93\u5165<\/span>Lithology Input<\/span><\/span>\n\u2192<\/span>\n\u641c\u7d22\u7b97\u6cd5<\/span>Search Algorithm<\/span><\/span>\n\u2192<\/span>\n\u5730\u5c42\u6392\u5e8f<\/span>Stratigraphic Ordering<\/span><\/span>\n\u2192<\/span>\n\u8de8\u5b54\u5bf9\u6bd4<\/span>Cross-hole Correlation<\/span><\/span>\n\u2192<\/span>\n\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Uncertainty Quantification<\/span><\/span>\n<\/div>\n\n<ul>\n<li>\u7cfb\u7edf\u641c\u7d22\u7b97\u6cd5\uff1a<\/strong>\u901a\u8fc7\u641c\u7d22\u7b97\u6cd5\u7cfb\u7edf\u63a2\u7d22\u6240\u6709\u5730\u8d28\u5408\u7406\u7684\u5730\u5c42\u6392\u5e8f\u65b9\u6848\uff0c\u800c\u975e\u5355\u4e00\u786e\u5b9a\u6027\u89e3<\/span>\nSystematic Search:<\/strong> Explores all geologically plausible stratigraphic orderings via search algorithms, not a single deterministic solution<\/span><\/li>\n<li>\u5730\u8d28\u5408\u7406\u6027\u7ea6\u675f\uff1a<\/strong>\u5f15\u5165\u5730\u8d28\u89c4\u5219\u7ea6\u675f\u641c\u7d22\u7a7a\u95f4\uff0c\u5982\u5730\u5c42\u53e0\u8986\u5f8b\u3001\u5ca9\u6027\u8fc7\u6e21\u6982\u7387\u7b49<\/span>\nGeological Plausibility Constraints:<\/strong> Introduces geological rules (superposition law, lithology transition probabilities) to constrain search space<\/span><\/li>\n<li>\u8de8\u94bb\u5b54\u5bf9\u6bd4\uff1a<\/strong>\u7ed3\u5408\u591a\u94bb\u5b54\u95f4\u5730\u5c42\u5355\u5143\u7684\u62d3\u6251\u5173\u7cfb\u8fdb\u884c\u5bf9\u6bd4\u9a8c\u8bc1\uff0c\u589e\u5f3a\u89e3\u91ca\u53ef\u9760\u6027<\/span>\nCross-hole Correlation:<\/strong> Correlates stratigraphic units across multiple drillholes using topological relationships to enhance reliability<\/span><\/li>\n<li>\u4e0d\u786e\u5b9a\u6027\u91cf\u5316\uff1a<\/strong>\u8f93\u51fa\u591a\u4e2a\u5730\u8d28\u4e00\u81f4\u6027\u89e3\u53ca\u5176\u6982\u7387\u5206\u5e03\uff0c\u4e3a\u5efa\u6a21\u63d0\u4f9b\u98ce\u9669\u4f9d\u636e<\/span>\nUncertainty Quantification:<\/strong> Outputs multiple geologically consistent solutions with probability distributions, providing risk basis for modeling<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Uncertainty Quantification<\/span><\/span>\n\u8de8\u5b54\u62d3\u6251<\/span>Cross-hole Topology<\/span><\/span>\nOpen Source<\/span>\u5f00\u6e90\u5b9e\u73b0<\/span><\/span>\n\n<ul>\n<li>\u5c06\u5730\u5c42\u89e3\u91ca\u4ece\u786e\u5b9a\u6027\u95ee\u9898\u91cd\u65b0\u5b9a\u4e49\u4e3a\u6982\u7387\u63a8\u65ad\u95ee\u9898\uff0c\u9996\u6b21\u5b9e\u73b0\u5168\u81ea\u52a8\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>\nRedefines stratigraphic interpretation from a deterministic to a probabilistic inference problem, achieving fully automated uncertainty quantification<\/span><\/li>\n<li>\u63d0\u51fa\u8de8\u94bb\u5b54\u5730\u5c42\u5355\u5143\u62d3\u6251\u5173\u7cfb\u5bf9\u6bd4\u65b9\u6cd5\uff0c\u5c06\u5355\u5b54\u89e3\u91ca\u7684\u53ef\u9760\u6027\u63d0\u5347\u5230\u591a\u5b54\u8054\u5408\u7ea6\u675f\u5c42\u9762<\/span>\nProposes cross-hole stratigraphic unit topological correlation, elevating single-hole interpretation reliability to multi-hole joint constraint level<\/span><\/li>\n<li>\u5f00\u6e90\u7b97\u6cd5litho2strat 1.0\u53ef\u76f4\u63a5\u96c6\u6210\u5230\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u5de5\u4f5c\u6d41\u4e2d<\/span>\nOpen-source algorithm litho2strat 1.0 can be directly integrated into 3D geological modeling workflows<\/span><\/li>\n<\/ul>\n<\/div>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">Optimizing Text Recognition in Borehole Log Images Using a Multi-Layout Adjustment Voting Mechanism<\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1aApplied Sciences (MDPI, 2025)<\/span>\nJournal: Applied Sciences (MDPI, 2025)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u94bb\u5b54\u67f1\u72b6\u56fe\u56fe\u50cf\u7684\u6587\u5b57\u8bc6\u522b\u51c6\u786e\u7387\u53d7\u9650\u4e8e\u590d\u6742\u5e03\u5c40\uff08\u591a\u680f\u3001\u65cb\u8f6c\u6587\u5b57\u3001\u8868\u683c\u7ebf\u5e72\u6270\u7b49\uff09\uff0c\u4f20\u7edfOCR\u76f4\u63a5\u5e94\u7528\u6548\u679c\u4e0d\u4f73\u3002<\/span>\nText recognition accuracy in borehole log images is limited by complex layouts (multi-column, rotated text, table line interference, etc.), where traditional OCR performs poorly.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<div class=\"method-flow\">\n\u56fe\u50cf\u8f93\u5165<\/span>Image Input<\/span><\/span>\n\u2192<\/span>\n\u591a\u5e03\u5c40\u8c03\u6574<\/span>Multi-layout Adjustment<\/span><\/span>\n\u2192<\/span>\n\u591a\u6b21OCR<\/span>Multiple OCR Runs<\/span><\/span>\n\u2192<\/span>\n\u6295\u7968\u6574\u5408<\/span>Voting Integration<\/span><\/span>\n\u2192<\/span>\n\u6700\u4f18\u7ed3\u679c<\/span>Optimal Result<\/span><\/span>\n<\/div>\n\n<ul>\n<li>\u591a\u5e03\u5c40\u8c03\u6574\u7b56\u7565\uff1a<\/strong>\u5bf9\u539f\u59cb\u56fe\u50cf\u8fdb\u884c\u591a\u79cd\u5e03\u5c40\u53d8\u6362\uff08\u5982\u5206\u680f\u5207\u5272\u3001\u65cb\u8f6c\u6821\u6b63\u3001\u5bf9\u6bd4\u5ea6\u589e\u5f3a\u7b49\uff09\uff0c\u751f\u6210\u591a\u4e2a\u5019\u9009\u8f93\u5165<\/span>\nMulti-layout Adjustment:<\/strong> Applies various layout transformations (column splitting, rotation correction, contrast enhancement) to generate multiple candidate inputs<\/span><\/li>\n<li>\u591aOCR\u7ed3\u679c\u751f\u6210\uff1a<\/strong>\u6bcf\u4e2a\u5019\u9009\u5e03\u5c40\u72ec\u7acb\u8fd0\u884cOCR\uff0c\u4ea7\u751f\u591a\u4e2a\u8bc6\u522b\u7ed3\u679c<\/span>\nMultiple OCR Results:<\/strong> Each candidate layout runs OCR independently, producing multiple recognition results<\/span><\/li>\n<li>\u6295\u7968\u673a\u5236\u6574\u5408\uff1a<\/strong>\u8bbe\u8ba1\u6295\u7968\u7b97\u6cd5\u5bf9\u591a\u8defOCR\u7ed3\u679c\u8fdb\u884c\u7f6e\u4fe1\u5ea6\u52a0\u6743\u6574\u5408\uff0c\u8f93\u51fa\u6700\u4f18\u8bc6\u522b\u7ed3\u679c<\/span>\nVoting Integration:<\/strong> Designed voting algorithm performs confidence-weighted integration of multi-path OCR results to output optimal recognition<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u591a\u5e03\u5c40\u6295\u7968<\/span>Multi-layout Voting<\/span><\/span>\n\u5de5\u7a0b\u6570\u636e\u96c6<\/span>Engineering Dataset<\/span><\/span>\n\n<ul>\n<li>\u63d0\u51fa”\u591a\u5e03\u5c40\u8c03\u6574+\u6295\u7968”\u7684\u65b0\u8303\u5f0f\uff0c\u7a81\u7834\u4e86\u5355\u4e00\u8def\u5f84OCR\u7684\u6027\u80fd\u74f6\u9888<\/span>\nProposes a “multi-layout adjustment + voting” paradigm, breaking through the performance bottleneck of single-path OCR<\/span><\/li>\n<li>\u5728\u5927\u5c39\u683c\u5e84\u548c\u4e01\u5bb6\u5c71\u4e24\u4e2a\u771f\u5b9e\u5730\u8d28\u6570\u636e\u96c6\u4e0a\u9a8c\u8bc1\uff0cF1\u503c\u5206\u522b\u8fbe\u523097.96%\u548c94.36%<\/span>\nValidated on two real geological datasets (Dayingezhuang and Dingjiashan), achieving F1 scores of 97.96% and 94.36% respectively<\/span><\/li>\n<li>\u65b9\u6cd5\u901a\u7528\u6027\u5f3a\uff0c\u53ef\u63a8\u5e7f\u5230\u5176\u4ed6\u590d\u6742\u8868\u683c\u6587\u6863\u7684OCR\u573a\u666f<\/span>\nMethod is generalizable and can be extended to other complex table document OCR scenarios<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box success\">\n\u5b9e\u9a8c\u7ed3\u679c\uff1a<\/span>Experimental Results:<\/span><\/strong><br\/>\n\u5927\u5c39\u683c\u5e84\u6570\u636e\u96c6 F1 = 97.96%<\/mark> | \u4e01\u5bb6\u5c71\u6570\u636e\u96c6 F1 = 94.36%<\/mark><\/span>\nDayingezhuang F1 = 97.96%<\/mark> | Dingjiashan F1 = 94.36%<\/mark><\/span>\n<\/div>\n<\/div>\n\n<\/section>\n\n<section class=\"section\" id=\"chinese\">\n\n<div class=\"section-icon\">\ud83c\udde8\ud83c\uddf3<\/div>\n<h2>\u4e2d\u6587\u6838\u5fc3\u7814\u7a76<\/span>Chinese Core Research<\/span><\/h2>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">\u6cb3\u5317\u77011\u223650 000\u6c99\u6d41\u6cb3\u5e45\u7b2c\u56db\u7cfb\u94bb\u5b54\u6570\u636e\u5e93<\/span>Quaternary Drilling Database of Shaliuhe Sheet, Hebei Province (1:50,000)<\/span><\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1a\u300a\u4e2d\u56fd\u5730\u8d28\u300b(2020)<\/span>\nJournal: Geology in China (2020)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u533a\u57df\u5730\u8d28\u8c03\u67e5\u4ea7\u751f\u5927\u91cf\u94bb\u5b54\u6570\u636e\uff0c\u4f46\u7f3a\u4e4f\u7cfb\u7edf\u5316\u7684\u6570\u636e\u5e93\u5efa\u8bbe\u65b9\u6cd5\uff0c\u6570\u636e\u8d28\u91cf\u53c2\u5dee\u4e0d\u9f50\uff0c\u96be\u4ee5\u652f\u6491\u540e\u7eed\u7684\u4e09\u7ef4\u5730\u8d28\u5efa\u6a21\u548c\u8d44\u6e90\u5171\u4eab\u3002<\/span>\nRegional geological surveys produce large amounts of drillhole data, but systematic database construction methods are lacking, data quality is inconsistent, and supporting 3D geological modeling and resource sharing is difficult.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<ul>\n<li>\u591a\u5c42\u6570\u636e\u6a21\u578b\uff1a<\/strong>\u5efa\u7acb1\u4e2a\u94bb\u5b54\u57fa\u672c\u4fe1\u606f\u56fe\u5c42 + 5\u4e2a\u94bb\u5b54\u7f16\u5f55\u6570\u636e\u5e93 + 5\u4e2a\u94bb\u5b54\u8bbe\u8ba1\u6570\u636e\u5e93 + 5\u5f20\u94bb\u5b54\u67f1\u72b6\u56fe\u7684\u5206\u5c42\u7ec4\u7ec7\u7ed3\u6784<\/span>\nMulti-layer Data Model:<\/strong> Establishes a hierarchical structure: 1 basic borehole info layer + 5 drilling log databases + 5 drilling design databases + 5 borehole columnar diagrams<\/span><\/li>\n<li>\u6570\u636e\u8d28\u91cf\u63a7\u5236\u4f53\u7cfb\uff1a<\/strong>\u4ece\u6570\u636e\u6765\u6e90\u3001\u5904\u7406\u65b9\u6cd5\u3001\u5c5e\u6027\u7ed3\u6784\u4e09\u4e2a\u7ef4\u5ea6\u5efa\u7acb\u8d28\u91cf\u63a7\u5236\u6d41\u7a0b<\/span>\nData Quality Control System:<\/strong> Establishes QC processes from three dimensions: data source, processing method, and attribute structure<\/span><\/li>\n<li>\u6807\u51c6\u5316\u5c5e\u6027\u7ed3\u6784\uff1a<\/strong>\u7edf\u4e00\u94bb\u5b54\u7f16\u53f7\u3001\u5750\u6807\u3001\u5b54\u53e3\u6807\u9ad8\u3001\u94bb\u673a\u7c7b\u578b\u3001\u5f00\u7ec8\u5b54\u65e5\u671f\u7b49\u5b57\u6bb5\u89c4\u8303<\/span>\nStandardized Attribute Structure:<\/strong> Unifies field specifications for borehole ID, coordinates, collar elevation, rig type, start\/end dates, etc.<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u6570\u636e\u5e93\u4f53\u7cfb<\/span>Database System<\/span><\/span>\n\u8d28\u63a7\u6d41\u7a0b<\/span>QC Process<\/span><\/span>\n\n<ul>\n<li>\u9996\u6b21\u7cfb\u7edf\u9610\u8ff0\u4e861:5\u4e07\u6bd4\u4f8b\u5c3a\u7b2c\u56db\u7cfb\u94bb\u5b54\u6570\u636e\u5e93\u7684\u5b8c\u6574\u5efa\u8bbe\u65b9\u6cd5<\/span>\nFirst systematic description of complete construction methods for 1:50,000 scale Quaternary borehole databases<\/span><\/li>\n<li>\u63d0\u51fa\u7684\u5206\u5c42\u6570\u636e\u7ec4\u7ec7\u7ed3\u6784\u517c\u987e\u4e86\u539f\u59cb\u7f16\u5f55\u6570\u636e\u3001\u8bbe\u8ba1\u6570\u636e\u548c\u6210\u679c\u56fe\u4ef6\u7684\u7ba1\u7406\u9700\u6c42<\/span>\nThe proposed hierarchical data organization accommodates management needs for original log data, design data, and output maps<\/span><\/li>\n<\/ul>\n<\/div>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">\u8de8\u5e73\u53f0\u94bb\u5b54\u67f1\u72b6\u56fe\u7684\u81ea\u52a8\u751f\u6210<\/span>Cross-Platform Automatic Generation of Borehole Columnar Diagrams<\/span><\/div>\n<div class=\"paper-meta\">\n\u671f\u520a\uff1a\u300a\u5ca9\u571f\u5de5\u7a0b\u6280\u672f\u300b(2016)<\/span>\nJournal: Geotechnical Engineering Technique (2016)<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u4f20\u7edf\u94bb\u5b54\u67f1\u72b6\u56fe\u4e0e\u7279\u5b9aCAD\u5e73\u53f0\u5f3a\u7ed1\u5b9a\uff0c\u4e0d\u540c\u5355\u4f4d\u4f7f\u7528\u4e0d\u540c\u8f6f\u4ef6\u5bfc\u81f4\u56fe\u4ef6\u65e0\u6cd5\u4e92\u901a\uff0c\u6570\u636e\u590d\u7528\u56f0\u96be\u3002<\/span>\nTraditional borehole columnar diagrams are tightly bound to specific CAD platforms. Different organizations using different software prevents interoperability and data reuse.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Methodology<\/span><\/h4>\n<ul>\n<li>\u901a\u7528\u6570\u636e\u6a21\u578b\uff1a<\/strong>\u8bbe\u8ba1\u72ec\u7acb\u4e8e\u4efb\u4f55\u5e73\u53f0\u7684\u94bb\u5b54\u901a\u7528\u6570\u636e\u6a21\u578b\uff0c\u5c06\u5ca9\u6027\u3001\u6df1\u5ea6\u3001\u63cf\u8ff0\u7b49\u8981\u7d20\u62bd\u8c61\u4e3a\u5e73\u53f0\u65e0\u5173\u7684\u6570\u636e\u7ed3\u6784<\/span>\nUniversal Data Model:<\/strong> Designs a platform-independent universal borehole data model, abstracting lithology, depth, descriptions into platform-agnostic data structures<\/span><\/li>\n<li>\u5ca9\u6027\u82b1\u7eb9\u81ea\u52a8\u586b\u5145\u7b97\u6cd5\uff1a<\/strong>\u6839\u636e\u5ca9\u6027\u7f16\u7801\u81ea\u52a8\u5339\u914d\u6807\u51c6\u82b1\u7eb9\u5e93\uff0c\u6309\u6df1\u5ea6\u95f4\u9694\u81ea\u52a8\u586b\u5145\u5ca9\u6027\u67f1<\/span>\nAutomatic Lithology Pattern Fill:<\/strong> Automatically matches standard pattern libraries by lithology code and fills lithology columns by depth interval<\/span><\/li>\n<li>\u77e2\u91cf\u56fe\u5143\u751f\u6210\uff1a<\/strong>\u751f\u6210\u72ec\u7acb\u4e8e\u5e73\u53f0\u7684\u77e2\u91cf\u56fe\u5143\u6570\u636e\u96c6\uff0c\u53ef\u88ab\u4efb\u4f55\u652f\u6301\u77e2\u91cf\u7684\u8f6f\u4ef6\u6e32\u67d3<\/span>\nVector Primitive Generation:<\/strong> Generates platform-independent vector primitive datasets renderable by any vector-supporting software<\/span><\/li>\n<\/ul>\n\n<h4>\u521b\u65b0\u70b9<\/span>Innovations<\/span><\/h4>\n\u8de8\u5e73\u53f0<\/span>Cross-Platform<\/span><\/span>\n\u81ea\u52a8\u6210\u56fe<\/span>Auto Plotting<\/span><\/span>\n\n<ul>\n<li>\u63d0\u51fa”\u4e00\u6b21\u751f\u6210\uff0c\u5230\u5904\u4f7f\u7528”\u7684\u8de8\u5e73\u53f0\u7406\u5ff5\uff0c\u7a81\u7834\u4e86CAD\u5e73\u53f0\u9501\u5b9a<\/span>\nProposes the “generate once, use everywhere” cross-platform concept, breaking through CAD platform lock-in<\/span><\/li>\n<li>\u5ca9\u6027\u82b1\u7eb9\u81ea\u52a8\u586b\u5145\u7b97\u6cd5\u89e3\u51b3\u4e86\u5927\u91cf\u94bb\u5b54\u6279\u91cf\u6210\u56fe\u7684\u5173\u952e\u6280\u672f\u74f6\u9888<\/span>\nThe automatic lithology pattern fill algorithm solves the key technical bottleneck for batch plotting of large numbers of boreholes<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box info\">\n\u6838\u5fc3\u4ef7\u503c\uff1a<\/span>Core Value:<\/span><\/strong>\n\u5c06\u6570\u636e\u4e0e\u8868\u73b0\u5206\u79bb\uff0c\u6570\u636e\u5b58\u50a8\u4e3a\u5e73\u53f0\u65e0\u5173\u7684\u4e2d\u95f4\u683c\u5f0f\uff0c\u56fe\u4ef6\u6e32\u67d3\u4ea4\u7531\u76ee\u6807\u5e73\u53f0\u5b8c\u6210\uff0c\u5b9e\u73b0\u4e86\u771f\u6b63\u7684\u8de8\u5e73\u53f0\u5e94\u7528\u3002<\/span>\nSeparates data from presentation: data is stored in a platform-independent intermediate format, and rendering is delegated to the target platform, achieving true cross-platform applicability.<\/span>\n<\/div>\n<\/div>\n\n<\/section>\n\n<section class=\"section\" id=\"conference\">\n\n<div class=\"section-icon\">\ud83d\udcdd<\/div>\n<h2>\u4f1a\u8bae\u8bba\u6587<\/span>Conference Papers<\/span><\/h2>\n\n<div class=\"paper-card\">\n<div class=\"paper-title\">Digitizing Drilling Logs – Challenges of typewritten forms<\/div>\n<div class=\"paper-meta\">\n\u4f1a\u8bae\uff1aGI\u4f1a\u8bae\u8bba\u6587\u96c6<\/span>\nConference: GI Conference Proceedings<\/span>\n<\/div>\n\n<h4>\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/span>Background & Problem<\/span><\/h4>\n\u5386\u53f2\u94bb\u5b54\u8bb0\u5f55\u5927\u91cf\u4f7f\u7528\u6253\u5b57\u673a\u586b\u5199\uff0c\u8fd9\u4e9b\u7eb8\u8d28\/\u626b\u63cf\u4ef6\u8868\u683c\u7684\u6570\u5b57\u5316\u9762\u4e34\u72ec\u7279\u6311\u6218\uff0c\u4e0e\u5370\u5237\u4f53\u6216\u624b\u5199\u4f53\u6587\u6863\u6709\u672c\u8d28\u4e0d\u540c\u3002<\/span>\nHistorical drilling logs were largely typewritten. Digitizing these paper\/scanned forms faces unique challenges fundamentally different from printed or handwritten documents.<\/span><\/p>\n\n<h4>\u6838\u5fc3\u65b9\u6cd5\u4e0e\u6280\u672f\u8def\u5f84<\/span>Core Methods & Technical Paths<\/span><\/h4>\n\u8bba\u6587\u5206\u6790\u4e86\u4e09\u5927\u6280\u672f\u8def\u5f84\uff1a<\/span>\nThe paper analyzes three technical paths:<\/span><\/p>\n\n<ol>\n<li>\u8868\u683c\u8bc6\u522b\u8def\u5f84<\/span>Table Recognition Path<\/span><\/strong><br\/>\n\u8bc6\u522b\u8868\u683c\u7ebf\u7ed3\u6784 \u2192 \u5b9a\u4f4d\u5355\u5143\u683c \u2192 \u63d0\u53d6\u5355\u5143\u683c\u5185\u6587\u5b57<\/span>\nRecognize table line structure \u2192 Locate cells \u2192 Extract text within cells<\/span><\/li>\n<li>\u8bed\u4e49\u5173\u8054\u8def\u5f84<\/span>Semantic Association Path<\/span><\/strong><br\/>\n\u8bc6\u522b\u5b57\u6bb5\u6807\u7b7e \u2192 \u5173\u8054\u90bb\u8fd1\u6570\u503c \u2192 \u6784\u5efa\u952e\u503c\u5bf9<\/span>\nRecognize field labels \u2192 Associate nearby values \u2192 Build key-value pairs<\/span><\/li>\n<li>\u6570\u503c\u63d0\u53d6\u8def\u5f84<\/span>Numerical Extraction Path<\/span><\/strong><br\/>\n\u9488\u5bf9\u6df1\u5ea6\u3001\u539a\u5ea6\u3001\u767e\u5206\u6bd4\u7b49\u6570\u503c\u578b\u5b57\u6bb5\u7684\u7279\u6b8a\u63d0\u53d6\u7b56\u7565<\/span>\nSpecial extraction strategies for numerical fields like depth, thickness, percentages<\/span><\/li>\n<\/ol>\n\n<h4>\u5173\u952e\u6311\u6218<\/span>Key Challenges<\/span><\/h4>\n<ul>\n<li>\u6253\u5b57\u673a\u5b57\u4f53\u7b49\u5bbd\u7279\u6027\u5bfc\u81f4\u7684\u5217\u5bf9\u9f50\u6b67\u4e49<\/span>\nColumn alignment ambiguity caused by monospace typewriter fonts<\/span><\/li>\n<li>\u590d\u5199\u7eb8\/\u78b3\u5e26\u9020\u6210\u7684\u5b57\u7b26\u7c98\u8fde\u548c\u6a21\u7cca<\/span>\nCharacter adhesion and blurring from carbon paper\/ribbon<\/span><\/li>\n<li>\u8868\u683c\u7ebf\u7f3a\u5931\u6216\u65ad\u88c2\u5bfc\u81f4\u7684\u7ed3\u6784\u8bc6\u522b\u56f0\u96be<\/span>\nStructural recognition difficulty due to missing or broken table lines<\/span><\/li>\n<li>\u8bed\u4e49\u5173\u8054\u9519\u8bef\uff08\u5982\u6df1\u5ea6\u503c\u4e0e\u5ca9\u6027\u63cf\u8ff0\u9519\u914d\uff09<\/span>\nSemantic association errors (e.g., depth values mismatched with lithology descriptions)<\/span><\/li>\n<\/ul>\n\n<div class=\"tip-box warn\">\n\u5b66\u672f\u4ef7\u503c\uff1a<\/span>Academic Value:<\/span><\/strong>\n\u8be5\u6587\u662f\u5c11\u6570\u4e13\u95e8\u8ba8\u8bba\u6253\u5b57\u673a\u586b\u5199\u8868\u683c\u6570\u5b57\u5316\u6311\u6218\u7684\u6587\u732e\uff0c\u5bf9\u5904\u740620\u4e16\u7eaa\u5386\u53f2\u94bb\u5b54\u6863\u6848\u5177\u6709\u91cd\u8981\u53c2\u8003\u4ef7\u503c\u3002<\/span>\nThis is one of few papers specifically addressing typewritten form digitization challenges, providing important reference for processing 20th-century historical drilling archives.<\/span>\n<\/div>\n<\/div>\n\n<\/section>\n\n<section class=\"section\" id=\"comparison\">\n\n<div class=\"section-icon\">\u2696\ufe0f<\/div>\n<h2>\u65b9\u6cd5\u5bf9\u6bd4\u4e0e\u521b\u65b0\u603b\u7ed3<\/span>Method Comparison & Innovation Summary<\/span><\/h2>\n\n<h3>\u8bba\u6587\u65b9\u6cd5\u7279\u5f81\u5bf9\u6bd4<\/span>Paper Method Characteristics Comparison<\/span><\/h3>\n\n<div class=\"table-wrap\">\n<table>\n<thead>\n<tr>\n<th>\u8bba\u6587<\/span>Paper<\/span><\/th>\n<th>\u6838\u5fc3\u65b9\u6cd5<\/span>Core Method<\/span><\/th>\n<th>\u6280\u672f\u8def\u7ebf<\/span>Technical Route<\/span><\/th>\n<th>\u4e3b\u8981\u521b\u65b0<\/span>Main Innovation<\/span><\/th>\n<th>\u8bc4\u4ef7\u6307\u6807<\/span>Evaluation Metric<\/span><\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>dh2loop 1.0<\/td>\n<td>\u8bcd\u5e93+\u6a21\u7cca\u5339\u914d<\/span>Lexicon + Fuzzy Matching<\/span><\/td>\n<td>NLP+\u89c4\u5219<\/span>NLP + Rules<\/span><\/td>\n<td>\u591a\u7ea7\u5ca9\u6027\u5206\u7c7b\u5f00\u6e90\u5e93<\/span>Multi-level lithology classification open-source library<\/span><\/td>\n<td>\u63d0\u53d6\u738793%\/86%\/34%<\/span>Extraction rate 93%\/86%\/34%<\/span><\/td>\n<\/tr>\n<tr>\n<td>litho2strat 1.0<\/td>\n<td>\u641c\u7d22\u7b97\u6cd5+\u8de8\u5b54\u5bf9\u6bd4<\/span>Search + Cross-hole Correlation<\/span><\/td>\n<td>\u7ea6\u675f\u641c\u7d22+\u62d3\u6251\u5206\u6790<\/span>Constrained Search + Topology<\/span><\/td>\n<td>\u5730\u5c42\u89e3\u91ca\u4e0d\u786e\u5b9a\u6027\u91cf\u5316<\/span>Stratigraphic interpretation uncertainty quantification<\/span><\/td>\n<td>\u5730\u8d28\u4e00\u81f4\u6027\u6982\u7387<\/span>Geological consistency probability<\/span><\/td>\n<\/tr>\n<tr>\n<td>\u591a\u5e03\u5c40\u6295\u7968OCR<\/span>Multi-layout Voting OCR<\/span><\/td>\n<td>\u591a\u5e03\u5c40\u8c03\u6574+\u6295\u7968<\/span>Multi-layout + Voting<\/span><\/td>\n<td>\u56fe\u50cf\u5904\u7406+\u96c6\u6210\u5b66\u4e60<\/span>Image Processing + Ensemble<\/span><\/td>\n<td>\u590d\u6742\u5e03\u5c40OCR\u4f18\u5316\u8303\u5f0f<\/span>Complex layout OCR optimization paradigm<\/span><\/td>\n<td>F1=97.96%\/94.36%<\/span>F1=97.96%\/94.36%<\/span><\/td>\n<\/tr>\n<tr>\n<td>\u6c99\u6d41\u6cb3\u5e45\u6570\u636e\u5e93<\/span>Shaliuhe Database<\/span><\/td>\n<td>\u5206\u5c42\u6570\u636e\u6a21\u578b+\u8d28\u63a7<\/span>Hierarchical Model + QC<\/span><\/td>\n<td>\u6570\u636e\u5e93\u5de5\u7a0b<\/span>Database Engineering<\/span><\/td>\n<td>\u533a\u57df\u94bb\u5b54\u6570\u636e\u5e93\u4f53\u7cfb<\/span>Regional borehole database system<\/span><\/td>\n<td>\u6570\u636e\u5b8c\u6574\u6027<\/span>Data completeness<\/span><\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5e73\u53f0\u81ea\u52a8\u751f\u6210<\/span>Cross-platform Auto Generation<\/span><\/td>\n<td>\u901a\u7528\u6a21\u578b+\u77e2\u91cf\u751f\u6210<\/span>Universal Model + Vector Generation<\/span><\/td>\n<td>\u6570\u636e\u4e2d\u95f4\u4ef6<\/span>Data Middleware<\/span><\/td>\n<td>\u5e73\u53f0\u65e0\u5173\u7684\u6210\u56fe\u65b9\u6cd5<\/span>Platform-independent plotting<\/span><\/td>\n<td>\u8de8\u5e73\u53f0\u517c\u5bb9\u6027<\/span>Cross-platform compatibility<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n\n<h3>\u6280\u672f\u53d1\u5c55\u8d8b\u52bf<\/span>Technology Development Trends<\/span><\/h3>\n\n<div class=\"tip-box\">\n\u8d8b\u52bf\u4e00\uff1a\u4ece\u89c4\u5219\u9a71\u52a8\u5230\u6570\u636e\u9a71\u52a8<\/span>Trend 1: From Rule-driven to Data-driven<\/span><\/strong><br\/>\n\u65e9\u671f\u7814\u7a76\u4f9d\u8d56\u6b63\u5219\u8868\u8fbe\u5f0f\u548c\u89c4\u5219\u5339\u914d\uff08\u5982dh2loop\u7684\u8bcd\u5e93\u65b9\u6cd5\uff09\uff0c\u6700\u65b0\u7814\u7a76\u5f00\u59cb\u5f15\u5165XGBoost\u3001\u89c6\u89c9\u5927\u6a21\u578b\u7b49\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u3002<\/span>\nEarly research relied on regex and rule matching (e.g., dh2loop’s lexicon method); latest research introduces XGBoost, vision LLMs, and other ML approaches.<\/span>\n<\/div>\n\n<div class=\"tip-box\">\n\u8d8b\u52bf\u4e8c\uff1a\u4ece\u786e\u5b9a\u6027\u5230\u6982\u7387\u5316<\/span>Trend 2: From Deterministic to Probabilistic<\/span><\/strong><br\/>\nlitho2strat\u5c06\u5730\u5c42\u89e3\u91ca\u7684\u4e0d\u786e\u5b9a\u6027\u663e\u5f0f\u91cf\u5316\uff0c\u4ee3\u8868\u4e86\u4ece\u5355\u4e00\u6700\u4f18\u89e3\u5230\u6982\u7387\u89e3\u96c6\u7684\u65b9\u6cd5\u8bba\u8f6c\u53d8\u3002<\/span>\nlitho2strat explicitly quantifies stratigraphic interpretation uncertainty, representing a methodological shift from single optimal solutions to probabilistic solution sets.<\/span>\n<\/div>\n\n<div class=\"tip-box\">\n\u8d8b\u52bf\u4e09\uff1a\u4ece\u5355\u70b9\u5de5\u5177\u5230\u5f00\u6e90\u751f\u6001<\/span>Trend 3: From Point Tools to Open Ecosystems<\/span><\/strong><br\/>\ndh2loop\u3001litho2strat\u7b49\u53d1\u8868\u4e8eGMD\u7684\u8bba\u6587\u5747\u63d0\u4f9b\u5b8c\u6574\u5f00\u6e90\u5b9e\u73b0\uff0c\u5b66\u672f\u6210\u679c\u5411\u5de5\u7a0b\u5de5\u5177\u8f6c\u5316\u7684\u5468\u671f\u5927\u5e45\u7f29\u77ed\u3002<\/span>\nPapers published in GMD (dh2loop, litho2strat) provide complete open-source implementations, significantly shortening the cycle from academic results to engineering tools.<\/span>\n<\/div>\n\n<\/section>\n\n<div class=\"footnotes\">\n \u94bb\u5b54\u67f1\u72b6\u56fe\u8bba\u6587\u65b9\u6cd5\u5206\u6790 | \u57fa\u4e8e\u516c\u5f00\u5b66\u672f\u6587\u732e\u6574\u7406<\/span>Borehole Log Paper Analysis | Compiled from public academic literature<\/span><\/p>\n<\/div>\n\n<\/div>\n\n\n<script>\nfunction switchLanguage(lang) {\n document.documentElement.lang = lang;\n document.body.className = (lang === 'en') ? 'en-active' : '';\n \n var btnZh = document.getElementById('btn-zh');\n var btnEn = document.getElementById('btn-en');\n \n if (lang === 'zh') {\n btnZh.classList.add('active');\n btnEn.classList.remove('active');\n } else {\n btnEn.classList.add('active');\n btnZh.classList.remove('active');\n }\n}\n\ndocument.addEventListener('DOMContentLoaded', function() {\n switchLanguage('zh');\n});\n\ndocument.querySelectorAll('a[href^=\"#\"]').forEach(function(anchor) {\n anchor.addEventListener('click', function(e) {\n e.preventDefault();\n var target = document.querySelector(this.getAttribute('href'));\n if (target) {\n target.scrollIntoView({ behavior: 'smooth', block: 'start' });\n }\n });\n});\n<\/script>\n\n<style media=\"print\">\n .lang-bar { display: none !important; }\n section { page-break-inside: avoid; }\n<\/style>\n\n<\/body>\n<\/html>\n\n","protected":false},"excerpt":{"rendered":"\u94bb\u5b54\u67f1\u72b6\u56fe\u6570\u5b57\u5316\u8bba\u6587\u7efc\u8ff0Review of Borehole Log Digitization Papers […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11,9],"tags":[],"class_list":["post-290","post","type-post","status-publish","format-standard","hentry","category-cad","category-ai"],"blocksy_meta":[],"_links":{"self":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts\/290","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=290"}],"version-history":[{"count":1,"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts\/290\/revisions"}],"predecessor-version":[{"id":291,"href":"https:\/\/numsimlab.com\/index.php?rest_route=\/wp\/v2\/posts\/290\/revisions\/291"}],"wp:attachment":[{"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=290"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=290"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/numsimlab.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=290"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}