针对螺栓松动检测作业需求,提出了一种将深度学习目标检测技术与数字图像处理技术相结合的螺栓松动图像检测方法.将Faster R-CNN应用于螺栓目标检测,设计了螺栓图像预处理、特征提取以及螺栓松动判别算法,采集一定数量的螺栓图片集,验证螺栓松动图像检测方法的可行性和有效性.实验表明:该螺栓松动图像检测方法能够比较精确地定位和识别一幅图像中的螺栓,且可以有效地检测出螺栓是否松动及其松动程度.该图像检测方法可为螺栓松动故障诊断提供了一种新的解决思路和技术方法.
According to the requirements of bolt looseness detection,a bolt looseness image detection method combining deep learning object detection technology and digital image processing technology is proposed.Faster R-CNN is applied to bolt object de-tection,and bolt image preprocess,feature extraction and bolt looseness discrimination algorithms are designed.A certain number of bolt image sets are collected to verify the feasibility and effectiveness of bolt looseness image detection method.Experiments show that the bolt looseness image detection method can accurately locate and identify the bolt in an image,and can effectively detect whether the bolts are loose and the degree of looseness.The image detection method can provide a new solution and technical refer-ence for bolt loosening fault diagnosis.