Research On Image Matching Search Strategy And Fitness Function

Image matching refers to finding the connections between two or more images which collected from the same place. At present, the technology of image matching is widely used in our daily life, for example, artificial intelligence, computer science, and information processing etc. It has become one of an essential task in the field of the digital image processing.By analysis of the drawbacks of the traditional global matching and the local matching strategy, such as the global minimum mean square error (MSE), the normalized minimum mean square error (NMSE), the normalized cross correlation coefficient, histogram and edge, corner based matching algorithm. According to simulate the process of the target moves from far to near, when the distance was far away, only the contour, shape and rough information was available to distinguish the aim target from others. While as the distance becomes more and more near, the texture, edge, corner and other details information will gradually appear which gives a more detailed description of the target. A new method by combination of the advantages of global and local features was proposed in this paper which shows robust to translation, rotation and changes on viewpoints. To speed up this algorithm, the genetic algorithm was used to search its global extremum, and a comparison of different fitness function was given to reveal the relation between speed and precision of this method and some practical proposes was cited in the end. The experimental results show that, our algorithm is superior to the several traditional matching algorithms.