基于DeepSORT和改进YOLOv5的煤矿井下钻杆计数方法

Drill Pipe Counting Method Based on DeepSORT and Improved YOLOv5 in Coal Mine Underground

针对煤矿井下钻杆计数存在精度较差、效率较低等问题,提出一种基于DeepSORT和改进YOLOv5的煤矿井下钻杆计数方法。首先,设计DR-C3模块,提高YOLOv5网络提取特征的能力;其次,引入GAM注意力机制,减少复杂背景的干扰;然后,通过CARAFE上采样算子扩大感受野;最后,结合DeepSORT算法对钻杆进行实时追踪计数。实验结果表明,改进后的YOLOv5 mAP@0.5提升了2.8%;钻杆计数平均精度达99.4%,检测速度达到93帧/s,计数精度高,满足实际需求。

A coal mine underground drill rod counting method based on DeepSORT and improved YOLOv5 is proposed to address the issues of poor accuracy and low efficiency in counting drill rods.Firstly,the DR-C3 module is designed,to enhancing the network′s feature extraction capability.Secondly,the GAM attention mechanism is introduced to reduce background interference.Then,the CARAFE upsampling operator is used to expand the receptive field.Finally,real-time tracking and counting of drill pipe in combination with DeepSORT algorithm.Experimental results show that the improved YOLOv5 achieves 2.8%increase in mAP@0.5,with an average precision of 99.4%in drill rod counting,and detection speed of 93 frame/s,and the counting accuracy is high,which meets the actual needs.