Research Of Shape Matching Based On CSS

Shape matching based on curvature scale space includes three main aspects. shape feature abstraction; improved corner detection algorithm; shape-matching algorithm. Firstly, the detection algorithms for edge feature and corner are discussed in this paper. next the multianalysis has contrasted the Moravec arithmetic operators, Harris arithmetic operators, SUSAN arithmetic operators and CSS arithmetic operators, from extraction efficiency, arithmetic operators stability, localization accuracy, Anti-noise,to carry on the analysis to compare to the operator with the test image to carry on the experimental analysis one by one. Secondly, adaptive corner detection algorithm based on direct curvature scale space is proposed based on classic CSS corner detection algorithm and improved algorithm. Under the direct curvature scale space, The improve algorithm uses curvature product with selected threshold to strengthen the corner detection and increases the number of the detecting true corners. Then check the angles of corner candidates in the dynamic region of support in order to eliminate falsely detected corners and use an adaptive curvature threshold to remove round corners from the initial list.The improved method not only improves the corner detection precision and stability, but also reduces the computational complexity. Finally, the simplified technology of corner detection value is applied in the shape-matching. To solve the problem of different sizes of the object shapes to be matched, the data resulting from improved corner detection are simplified and normalized at first. Then a matching value is evaluated which measures the similarity between different object shapes. Numerical experiments show that the proposed algorithm has a good performance both for corner detection and shape matching.The last chapter in the paper mainly carries on summary and forecast, summarizes the work of the paper, points out the good and bad points, proposes center of gravity of work in the future.