基于改进YOLOv5的煤岩图像检测识别研究  

Research on Coal Rock Image Detection and Recognition Based on Improved YOLOv5

作　　者：王建才[1] 李瑾 李志军[2] 史健婷[1] 唐垚锐 Wang Jiancai;Li Jin;Li Zhijun;Shi Jianting;Tang Yaorui(Heilongjiang University of Science and Technology,Harbin 150022,China;Wuzhou University,Wuzhou 543002,China)

机构地区：[1]黑龙江科技大学,哈尔滨150022 [2]梧州学院,广西梧州543002

出　　处：《煤矿机械》2022年第9期204-208,共5页Coal Mine Machinery

基　　金：黑龙江省大学生创新训练项目(202110219048);2020年度黑龙江省省属本科高校基本科研业务费项目(YJSCX2020-212HKD)。

摘　　要：针对机器学习方法实现煤岩图像识别存在的问题,利用YOLOv5网络开展煤岩图像的检测识别。将CBAM注意力机制和Transformer添加到YOLOv5网络模型中,改进传统YOLOv5网络检测精度不高的问题。结果显示,添加CBAM注意力机制的YOLOv5网络模型平均精度最高达到0.928。利用添加了CBAM注意力机制的YOLOv5网络模型对煤岩图像进行检测,将识别结果中的煤岩混合图像从样本中分类出来,用于后续的图像分割,减少直接进行数据分割的运行时间以及误差。In view of the problems of traditional machine learning in coal and rock image detection and recognition, the detection and classification of coal and rock images were carried out by using YOLOv5 network. CBAM attention mechanism and Transformer were added to the YOLOv5 network model to improve the problem of low detection accuracy of traditional YOLOv5 network. The results show that the average accuracy of YOLOv5 network model with CBAM attention mechanism is the highest,reaches 0.928. YOLOv5 with CBAM attention mechanism was used to detect and recognize coal and rock images. The coal rock mixture images in the recognition results were classified from the samples for subsequent image segmentation, which are used for subsequent image segmentation to reduce the running time and error of direct data segmentation.

关 键 词：机器学习 煤岩图像检测识别 YOLOv5 CBAM注意力机制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]