轮廓匹配方法改进及其在转子中的应用

Improvement of contour matching method and its application in rotors

汽车发动机转子装配误差容易造成零件磨损和效率降低.针对转子零件装配精度要求,提出一种基于轮廓与特征点结合的轮廓匹配改进方法.首先对转子图像进行滤波、二值化预处理;其次提取轮廓并利用轮廓Hu矩方法进行轮廓粗匹配;然后提取特征点,并在以特征点为中心、半径为16像素的圆形区域内生成4个特征描述子作为匹配特征,利用欧氏距离进行特征点匹配;最后通过找到的匹配对计算仿射变换矩阵,得到精匹配结果.实验结果表明,该方法在背景复杂、光照变化的条件下能够较精确地匹配轮廓,保证匹配精度在1像素内,且生成的特征描述子对数越多,匹配结果越精确.

Vehicle engine rotor assembly error is easy to cause parts friction and efficiency reduction.Aiming at the assembly precision requirements of rotor parts,a improvement of contour matching method based on the combination of contour and feature points was proposed.Firstly,the image is preprocessed by filtering and binarization.Secondly,contour matching was carried out by contour Hu moment method.Then the feature points were extracted,and four feature descriptors were generated in the circular region with the radius of 16 pixels and the feature points were matched by Euclidean distance.Finally,the affine transformation matrix is generated by finding the matching pair,and the precise matching result is obtained.The experimental results show that the proposed method can match the contour accurately under the condition of complex background and light change,and ensure the visual detection accuracy within one pixel.The more the number of feature descriptors generated,the more accurate the matching result.