Dark field image registration based on connection vector feature matching method

Based on the the positional relationship between damage points and characteristics of optical damage dark field image, a registration method using connection vector feature matching was designed. The method firstly preprocessed the reference image and the image to be registered, and then extracted the center coordinates of the contour of the damage point as the position value of the damage point. Then the connection vector of the damage points was constructed, the main direction was obtained and the connection vector feature in the main direction was calculated, the exact match was achieved through the fine matches. Finally, the affine transformation parameter was calculated by using the RANSAC algorithm. The method had rotational invariance, scale invariance and high registration accuracy. The computational efficiency and registration precision of this method and SIFT algorithm were compared and analyzed. The experimental results show the more effectiveness of the proposed method in dark field image.