江西科学2025，Vol.43Issue(5)：848-856,883,10.DOI:10.13990/j.issn1001-3679.2025.05.005
基于YOLOv8算法的地质产状语义识别研究
Research on Semantic Recognition of Geological Attitudes Based on YOLOv8 Algorithm
摘要
Abstract
The automated recognition of geologic attitude information in topographic and geo-logic maps is of great significance to enhance the value of traditional geologic map data.In this study,a dual database containing 20 082 dip annotations and 1 820 keypoint annotations was constructed for the visual characteristics of geologic attitude symbols,and an efficient detection model was developed based on the YOLOv8 algorithm framework.By comparing the performance of different scale models,YOLOv8l was selected as the dip symbol detec-tion model(with 99.35%precision,99.18%recall,99.44%for mAP50,and 98.64%for mAP50-95),and YOLOv8m-Pose was selected as the keypoint detection model(with 98.79%precision,99.49%recall,99.33%for mAP50 and 99.27%for mAP50-95).The method is applied to a 1∶200 000 geological map of Qinyuan,Shanxi province,successfully identifying 326 out of 353 attitudes(92.35%),and to a 1∶250 000 geological map of Walvis Bay,Namibia,extracting 353 out of 408 attitudes(86.52%).The method overcomes the bottleneck of digitizing traditional geological maps and realizes the precise positioning of geo-logical feature symbols and the simultaneous analysis of feature elements(strike,dip direc-tion,and dip angle).The research results significantly improve the information extraction efficiency of geological maps,provide high-precision data support for regional geological surveys,mineral resources exploration and other applications,and offer practical value for promoting the digital transformation of geological data.关键词
地质产状/YOLOv8/目标检测/关键点检测Key words
geologic attitude/YOLOv8/target detection/keypoint detection分类
社会科学引用本文复制引用
张雪,黄映聪,黎广荣,赵月,谢安..基于YOLOv8算法的地质产状语义识别研究[J].江西科学,2025,43(5):848-856,883,10.基金项目
国家自然科学基金项目(42430208) （42430208）
中广核铀业发展有限公司科研项目(019GNB2024C45PO9900041). （019GNB2024C45PO9900041）
