Image matching based on shape includes three main aspects:shape feature abstraction; shape similarity measurement; optimum matching search. Firstly, the detection algorithms for edge feature and corner are discussed in this paper. A fast corner detection algorithm is presented to extract corners quickly. A new feature detection algorithm is also described to effectively eliminate the influence of different lighting condition of images'adjacent surface in the feature detection. Secondly, a fast strategy and an improved partial Hausdorff distance are proposed based on Hausdorff distance to roughly and accurately measure the shape similarity respectively. Also a new genetic algorithm with the probabilities of crossover and mutation self-adaption based on fuzzy control is adopted for fast search of optimum matching. Finally, several composite models are given for different application demands and their matching results are verified through the experiments. |