改进YOLOv5的矿用输送带损伤检测方法  

Improved YOLOv5 mine conveyor belt damage detection method

作　　者：张舒 叶涛[1] 陈云 Zhang Shu;Ye Tao;Chen Yun

机构地区：[1]武汉理工大学机电工程学院,武汉430070

出　　处：《起重运输机械》2024年第2期31-36,共6页Hoisting and Conveying Machinery

摘　　要：文中针对目前输送带损伤检测大多是输送带撕裂且缺乏其他损伤类型研究的问题,提出一种改进YOLOv5的矿用输送带损伤检测方法。将SPD-Conv模块替换Conv模块中的卷积层,提升小目标的检测效果;在骨干特征网络与最后预测网络之前引入CBAM注意力机制,对重要的特征通道进行强化;最后,在YOLOv5的基础上引入高斯滤波器消除噪声干扰,提升算法目标检测效率。试验结果表明:改进后的YOLOv5目标检测网络在对输送带的撕裂、击穿、表面划伤、破损4种损伤类型的检测平均精度均值达92.3%,相较于YOLOv5算法提高了35.1%,检测速度达90帧/s,提高了20%,实现了对矿用输送带损伤的快速识别。Considering that the most damage detection of conveyor belts is about the tearing of conveyor belt and there is a lacking of other damage types,an improved mine conveyor belt damage detection method is proposed.SPD-Conv module is used to replace the Convolution layer in conv module to improve the detection effect of small targets;Introducing CBAM attention mechanism before backbone feature network and final prediction network to strengthen important feature channels;Finally,Gaussian filter is introduced to eliminate noise interference on the basis of YOLOv5,so as to improve the efficiency of target detection.The experimental results show that the average detection accuracy of the improved YOLOv5 target detection network for four types of damage of conveyor belt is 92.3%,which is 35.1%higher than that of YOLOv5 algorithm,and the detection speed is 90 frames/s,which is 20%higher,thus realizing the rapid identification of mine conveyor belt damage.

关 键 词：矿用输送带 损伤检测 注意力机制 高斯滤波器 YOLOv5 

分 类 号：TH222[机械工程—机械制造及自动化]