现代电子技术2025，Vol.48Issue(9)：1-7,7.DOI:10.16652/j.issn.1004-373x.2025.09.001
基于改进YOLOv8的高效岩屑实例分割算法
Improved YOLOv8 based efficient instance segmentation algorithm of cuttings
摘要
Abstract
In the field of complex lithological determination from rock cuttings,traditional methods face issues of low efficiency and insufficient accuracy.In view of this,a model based on the improved YOLOv8 algorithm is presented to significantly enhance the efficiency and accuracy of instance segmentation of rock cuttings generated during drilling.The method of grouped focus downsampling instead of the partial conventional downsampling is proposed,which maintains accuracy while reducing computational complexity.Additionally,the redesigned partially coupled segmentation head reduces the parameters significantly,and improves the lightweight of the model.The integration of a window attention mechanism in the feature fusion layer further enhances the ability to capture rock cutting features and the overall performance of the model.Experimental validation shows that the improved Rock-YOLO model achieves a mean average precision(mAP)of 87.6%on a rock cuttings sampling dataset,and an improvement of 3.9%over that of the original model.Its parameters are just 8.3×106,a decrease of 29.1%,and its computational burden is also reduced to 3.5×1010,a decrease of 17.5%.These improvements ensure accuracy in analysis while meeting the needs for rapid response in the field,which makes the proposed model highly suitable for practical lithological determination.The combination of lightweight and high accuracy offers new possibilities for the application of deep learning in the field of geology.关键词
实例分割/YOLOv8/轻量化/岩性判断/移动设备/部分卷积Key words
instance segmentation/YOLOv8/lightweight/lithological determination/mobile device/partial convolution分类
信息技术与安全科学引用本文复制引用
刘学彬,钟宝荣..基于改进YOLOv8的高效岩屑实例分割算法[J].现代电子技术,2025,48(9):1-7,7.基金项目
国家自然科学基金项目(62006028) （62006028）
湖北省自然科学基金项目(2023AFB909) （2023AFB909）
