基于机器视觉和深度学习的矿用电机换向器检测系统设计研究

Design of Mine Motor Commutator Detection System Based on Machine Vision and Deep Learning

借助神经网络深度学习的算法和机器视觉技术,设计一种矿用电机换向器缺陷无损的自动检测系统,以解决矿用电机换向器经常出现的几何误差和表面质量的问题。通过采集零件图形、计算滤波预处理、区域亚像素定位、模板匹配等系列技术手段,分别从零件的顶面、侧面、底部等位置完成电机换向器缺陷无损检测。在电机换向器检测实验结果中发现,该系统检测过程简单,相比人工检测效率能够提高70.6%,漏检率0%,误检率降低到1%,实现检测效率的大幅度提升,同时能测量不同尺寸换向器的多种技术参数,这为换向器的检测提供了一种设计思路,同时也为其他零部件质检提供了优化思路。

With the help of neural network deep learning algorithm and machine vision technology,an automatic defect-free detection platform for mining electric motor commutators is designed which is pro⁃posed to address the frequent problems of morphological and positional errors and surface quality of commu⁃tators.The defectless detection is carried out from various positions including the top,sides,bottom and oth⁃er positions.The basic idea is to complete the detection through a series of technical means such as collect⁃ing part graphics,computer filtering preprocessing,regional subpixel positioning,template matching,etc.The commutator detection experiment found that the system has a simple detection process,which,compared to manual detection,can improve the efficiency by 70.6%,with missed detection rate of 0%and a false detec⁃tion rate reduced to 1%,achieving a significant improvement in detection efficiency.At the same time,it can measure various technical parameters of commutators of different sizes,which provides a design idea for the detection of commutators and optimization ideas for the quality inspection of other components.