A New Approach of Template Matching and Localization Based on the Guidance of Feature Points

Abstract

Template matching is the most basic and commonly used matching method in image processing. On account of different selections in matching, there are mainly two approaches: gray information matching and feature matching. The former method is accurate, but might be easily influenced by interferences such as illumination and other noises. In addition, the method is limited by its high time complexity. For the latter method, the anti-interference ability is relatively strong due to the extraction of image feature. However, most of feature extraction are very complex, which result in long computation time. In this paper, a novel approach of template matching and object localization based on the feature points in the edges of image is proposed. Firstly, the intersection points of principal axis and vice axis with the image edges are selected as the feature points; Secondly, the transformation parameters between the points set of template and object image edges are computed by using the feature points gotten in last step; Then, the points set of template contour is transformed and the Hausdorff distance of the transformed template and target points set are calculated as the similarity measure; Lastly, the minimum Hausdorff distance is searched by changing the transformation parameters in their respective neighborhoods, and the parameters corresponding are required. The experiment results indicate that this approach has a good performance both on not processing speed and accuracy.