Research On Image Matching Algorithm In Solder Joint Inspection System

Image matching became a hot and indispensable topic in computer vision field. For the sake of more precisely making out the target, it is mainly used to match two or several images that are gained from multi-sensor, at several time, from multi-view. It has become a fundamental problem in the fields of computer vision and image processing. At the same time, it also became a challenging issue of multi-sensor information fusion, which possesses considerable importance to theory and practice.In this thesis, we researched the algorithm about image matching that is applied to the solder joint inspection system. This thesis proposed an accelerated matching method based on image gradient characteristic. First, artificial selection of image feature matching points reduced the time of the algorithm searching for feature points; Prewitt operator was used to extract feature points and the corresponding gradient value, which can make matching feature more refined. Designing database of image translation, rotation and scaling accelerated the speed of image gradient matching. The thesis improved the image gradient algorithm, which compares gradient cosine. The maximum of cosine is the corresponding best matching between templates and images, which can reduce the time of the least squares optimization; On this basis, the distance error between the feature points is optimized by least square method, to further improve the accuracy of matching. This thesis, based on image gradient matching feature, can strengthen the robustness of algorithm for image rotations, translation, scaling, brightness changing.