Research on the Fast Template Matching Method of Micro Parts
Jiayi Zhang, Zhiqiang Liu, Ying Liu · 2020
Through the analysis of the characteristics of high detection frequency, various parts shape, random position and angle in the feeding process of large batch of micro parts assembly, the specific requirements of matching recognition of micro parts are elaborated, such as fast matching speed and affine transformation functions of displacement, rotation and flip. Based on the analysis of the existing typical image matching algorithms, a hybrid multi-level acceleration method is proposed with NCC similarity evaluation algorithm. According to the overlap rate of contour area of the binary template before and after affine transformation, the acceleration initial parameters can be calculated. By calculating the relative proportion of sampling points in contour area of the binary template before and after sampling, the sampling step size of template can be computed. On this basis, the specific steps of cycle multi-level acceleration are described in detail. An example is given to show the convergence of each parameter in the template matching process. The comparison results of six typical matching methods show that the accelerated NCC matching method has the advantages of high similarity, small error, fast speed, good robustness to the effect of illumination and texture, and meet the requirements of template matching in micro parts assembly process.