基于机器视觉的带式输送机托辊故障检测  

Fault detection of belt conveyor roller based on machine vision

作　　者：苗长云[1] 郭盈辉 刘意 MIAO Chang-yun;GUO Ying-hui;LIU Yi(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China;Teaching Practice Training Center,Tiangong University,Tianjin 300387,China)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津工业大学教学实习训练中心,天津300387

出　　处：《天津工业大学学报》2023年第3期56-63,共8页Journal of Tiangong University

基　　金：国家自然科学基金资助项目(51274150);天津市重点研发计划项目(18YFJLCG00060);天津市重点研发计划科技支撑重点项目(18YFZCGX00930)。

摘　　要：针对带式输送机托辊故障检测中存在成本高、可靠性和准确性差的问题,提出了基于机器视觉的带式输送机托辊故障检测方法。该方法利用摄像头采集托辊图像序列,利用YOLOv4-Tiny目标检测算法提取托辊感兴趣区域(ROI),通过托辊分割算法提取托辊表面区域图像,采用帧间差分的相邻两帧托辊表面区域图像处理算法判断托辊是否转动,若检测到托辊处于转动状态,采用密集采样算法提取托辊表面区域图像特征点,采用金字塔LK稀疏光流法跟踪特征点,根据相邻两帧图像同一特征点的位置计算托辊转速,能够实现根据托辊转速检测托辊故障。实验结果表明:该方法的托辊不转动(卡转)检测准确率为97.5%,转速检测最大误差不超过3.6%,可用于托辊转速故障检测,具有检测成本低、可靠性和准确性高等优点。In view of the problems of high cost,poor reliability and poor accuracy in the fault detection of roller for belt conveyors,a fault detection method for roller of belt conveyors based on machine vision was proposed.In this method,the image sequence of roller was collected by camera,the region of interest(ROI)of roller was extracted by YOLOv4-Tiny target detection algorithm,and the image of roller surface was extracted by roller segmentation algorithm.The image processing algorithm for the surface area of the roller of two adjacent frames using the difference between frames determined whether the roller was rotating.If the roller was detected in the rotation state,the image feature points of the roller surface area were extracted by the dense sampling algorithm.The Pyramid LK sparse optical flow method was used to track the feature points,and the roller speed was calculated according to the position of the same feature point in the two adjacent frames.It could detect roller fault according to roller speed.The experimental result shows that the detection accuracy of the roller non-rotation(stuck rotation)of this method is 97.5%,and the maximum error of the speed detection is not more than 3.6%.It can be used for the fault detection of the roller speed,and has the advantages of low detection cost,high reliability and accuracy,etc.

关 键 词：带式输送机 托辊故障检测 机器视觉 转速检测 图像处理算法 

分 类 号：TD634.1[矿业工程—矿山机电]