The Research And Implementation Of Two-Dimension Fragmental Objects Matching Methods

Fragmental objects recovery technology is a important research topic in computer vision, image analysis and pattern recognition. It opens up a new areas of application of pattern recognition, with a wide range of practical value, and has been concerned by domestic and foreign scholars. Based on the study of traditional corner extraction algorithms and traditional matching algorithm based on the combination of pattern recognition, image processing, science, differential geometry and shape matching, and other relevant theory, matching methods, a planar irregular fragmental corner detection method and a fragmental objects matching method are brought forward fit for this task. The main task of this paper is as follows:Firstly, according to the basic theory on curve and the shape representation criteria, curvature descriptor is introduced to describe outlines. Based on the method analysis of the discrete curve curvature calculation, an approximate calculation method of the discrete curve curvature is given. The experiments show that the algorithm has strong anti-interference of the advantages of digital, to better reflect the profile of the local features, suitable for following corner extraction and debris matching work.Secondly, the interior and exterior corner detection methods are researched, aiming at the question that the result is impacted by the digital noise of the contour. Two methods of corner detection are given. One is based on the contour sharp degree. The other is based on the approximate curvature calculation. It is showed that the algorithms can reduce the influence of the contour digitalization effectively.Lastly, aiming at the reason that at present the calculating amount of fragmental objects matching methods is very large, based on the character points detection and by means of invariants of curve and the hausdorff distance, a fragment matching algorithm is given based on the matching segments which is guided by the corner matching. The algorithm can improve the efficiency of contour matching.