基于深度学习的输煤皮带故障视觉检测方法研究

为了更好地检测皮带跑偏、撕裂和异物干扰等严重影响皮带安全运行的故障状态，围绕相关问题产生的原因及检测方法开展深入研究，通过对纵/横向裂缝、异物的检测分析、实验，提高基于视觉的检测精度。提出基于Canny边缘检测算法的皮带跑偏检测算法；基于深度学习的横向与纵向撕裂检测，尤其对于裂缝与纵向纹理区分不明显情况，提出一种红光透射的判别方式；基于最小距离分类算法将识别异物转换为分类问题，利用机器学习的方法对样本进行训练并建立无异物阈值，通过提取特征，最后利用最小距离分类算法得到有无异物的结果。研究结果表明:提出的视觉检测系统可以实时高效地检测出输煤皮带常见的3种故障，可进一步保障运输系统安全运行。

Research on visual detection method for fault of coal conveyor belt based on deep learning

In order to better inspect the fault conditions that seriously affect the safe operation of the belt,such as belt deviation,tearing and foreign body interference,the in-depth research on the causes and detection methods of related problems was conducted,and the detection,analysis and experiment on the longitudinal/ transverse cracks and foreign body were carried out to improve the accuracy of vision-based detection.An detection algorithm of belt deviation based on the Canny edge detection algorithm was put forward.Based on the deep learning of longitudinal and transverse tearing detection,especially for the case where the distinction between cracks and longitudinal textures was not obvious,a judgment method of red light transmission was proposed.Based on the minimum distance classification algorithm,the identification of foreign body was converted into a classification problem.The samples were trained by the machine learning,and the threshold of without foreign body was established.By extracting the characteristics,the minimum distance classification algorithm was finally used to obtain the results of whether there were foreign body.The results showed that the proposed visual detection system could efficiently detect the three common faults of coal conveying belt in real time,which can further ensure the safe operation of the transportation system.