基于机器视觉的汽车零部件检测系统设计

Auto Parts Inspection System Based on Machine Vision

文中设计了一套基于机器视觉的汽车零部件检测系统,实现对零部件冲孔和字符的检测。在对冲孔进行检测时,先使用阈值分割方法,再利用像素点面积特征和圆度特征来提取冲孔区域,最终完成对冲孔直径的检测。在对零件字符进行检测时,使用了深度学习的方法,将自然场景下的文本检测方法应用到零部件的字符检测中,解决了字符遮挡、产品色差变化较大以及产品摆放倾斜造成的字符定位困难问题,并且无需显示加入字符分割,实现端到端的字符识别。经过现场的测试验证,该检测系统能够实现多种型号零件的检测,具有良好的检测效果。

In this paper,a set of auto parts detection system based on machine vision was designed to realize the detection of punching and characters of parts.In the detection of punching,firstly,used the threshold segmentation method,then used the pixel area and roundness features to extracted the punching area,finally completed the detection of punching diameter.In the process of character detection of parts,the method of deep learning was used and the text detection method in the natural scene was applied to the character detection of parts,which solved the problem of character location caused by character occlusion,large product color difference change and inclined product placement,realized end-to-end character recognition without adding character segmentation.The detection system can realize the detection of various types of parts by field test verification,and has a good detection effect.