A fast, Accurate and Robust Image Matching Algorithm
XIONG Xing-hua1, CHEN Ying2, QIAN Zeng-bo3
(1. Xi'an Research Institute of Surveying and Mapping, Xi'an 710054, China; 2. The Research Center of Remote Sensing Techniques, Tongji University, Shanghai 200092, China; 3. Institute of Surveying & Mapping Information Engineering University, Zhengzhou 450052, China)
Abstract£ºImage Matching is very pivotal technique in the application of remote sensing image whose quality determines the performances of the subsequent production straightforwardly, and the qualities of matching result self are determined by the speed, precision and reliability of image matching algorithm. A new fast, accurate and robust image matching algorithm integrating wavelet transform, genetic algorithm with least square matching(LSM) is discussed. The algorithm makes use of the advantages of the above three techniques fully. Wavelet transform is used to reduce the search data size, and genetic algorithm is used to optimize the search solution space, and LSM is used to obtain the subpixel matching result. Simultaneously, an improved genetic algorithm with adaptive operator probability is presented. The basic principles are that the crossover probability of genetic algorithm varies with the change of the Hamming distance between two-parent chromosomes selected and the mutation probability varies with the change of the fitness of individual chosen. The experiment results have shown that the algorithm discussed is superior to the traditional ones in matching performances distinctly.
Key words£ºimage matching; wavelet transform; genetic algorithm; least square matching