Abstract: Aiming at the shortcomings of low efficiency, low accuracy and high labor intensity in the detection and recognition of chassis frame shape defects on the assembly line, a frame defect detection and recognition method based on image registration is proposed. The standard image is obtained, and the shape matching template is established by the maximum between class variance method and morphology. The frame image to be tested is collected online, and the position and rotation angle of the frame image to be tested in the standard image are found by using the shape template matching algorithm. The frame image to be tested is registered so that it is in the same spatial position relationship with the standard frame image. The target image to be tested is matched with the standard library image by logical XOR, and the interference defects and impurities in the system are removed by morphological algorithm. After RGB component filtering, the image is overlapped and the transparency is adjusted to obtain the defect image. The research results show that this method can effectively extract the shape characteristics of the frame, and then obtain a variety of shape defects such as over riveting, missing riveting, porous, few holes, etc., with fast detection speed and good robustness.

Key words: Frame; Image registration; Shape template matching; MATLAB; Defect detection

摘要： 针对流水线上底盘车架外形缺陷检测和识别存在效率低、准确度不高、劳动强度大的 缺点，提出一种基于图像配准的车架缺陷检测与识别方法。获取标准图像，通过最大类间方 差法与形态学等操作建立形状匹配模板。在线采集待测车架图像，利用形状模板匹配算法查 找待测车架图像在标准图像中的位置与旋转角度，对待测车架图像进行配准，使其与标准车 架图像处于同一空间位置关系。通过逻辑异或把待测目标图像与标准库图像进行匹配运算， 采用形态学算法去除系统中干扰缺陷杂质。RGB 分量滤波后，进行图像重叠并调整透明度， 从而得到缺陷图。研究结果表明，该方法可有效提取车架外形特征，进而得到过铆、漏铆、 多孔、少孔等多种外形缺陷，其检测速度快，具有较好的鲁棒性。

关键词: 车架；图像配准；形状模板匹配；MATLAB；缺陷检测