An Image Matching Method Based On Crossing Points

Shape matching is a comparison between two images, and is widely used in military, commercialization and civilian. In recent years, research about image matching is a hot point. Algorithms proposed now can only be used in certain cases, and are not general. Image matching technique is a hard point, variational shape matching and part coverage problem are not solved. Today's research concerns on designing new algorithms and improving the usage's range and efficiency of algorithms present.Shape is important information of objects, and matching algorithms based on shape are important branches, which can be divided into algorithms based on outline and algorithms based on district. The former one works fast, but the resolving capability is low; while the latter one has good resolving capability, but is sensitive to noise.This paper do research on image matching, include contents as below:1) Review popular image matching techniques, and concerns on algorithms based on shape.2) Propose an algorithm of model construction based on crossing points character. It uses a group of concentric circles to cross with the edge, so the model stand for a range of deformation.3) With the help of the model, a matching method based on crossing point is dressed. It uses the number of crossing points and the distance between crossing points. It is unrelated with the position, size and rotation of image, and can recognize some shape change. When used to match natural object, this method works well.4) Based on algorithms above, an implement of the modeling and matching method is carried out.At last of this paper, according to the comparison to hausdorff image matching method, it proved to be better in speed and discrimination.