An Image-scene Matching Method For Varied Environment

Scene matching technology application field is very extensive when it passes through several dozens years development. For example, digital image processing, navigation and guidance, environmental research and so on are all based on scene matching technology. The generalized application has produced many different matching algorithms, each matching algorithm is suitable for specific domain. The paper mainly studies a algorithm which can run at environmental changes, including rotation angle, resolution ratio, contrast scale differences ,translation and noise.The paper detailed describes related knowledge of scene matching, it provides theoretical basis and reference parameters. According to characteristics, scene matching algorithms can divided into four categories,. The paper combined use two categories: gray based matching and characteristics based matching.In scene matching, rotation angle, resolution ratio, contrast scale differences ,translation and noise are the main causes of the false matching. The algorithm which the paper study out must overcome these distortions as possible. A new scene matching method was proposed for finding the matching point accurately. In this method, firstly, rejected the pixels in the image that are not in the maximum inscribed circle. Secondly, combined with Log-Polar Transformation which can transform the rotation and contrast scale differences in Cartesian coordinates on the image to translational changes in horizontal and vertical directions in log-polar coordinates. Finally, used the Fourier Transformation to take the amplitude value which had such trait as invariance to translation to match. Search strategy use Pyramid Hierarchical Algorithm, measure method use Removing Mean Normalized Product Correlation Algorithm. The paper detailed describes related knowledge of Log-Polar Transformation, Fourier Transformation, Pyramid Hierarchical Algorithm and Removing Mean Normalized Product Correlation Algorithm, and using theoretical and experiments verified the correctness of the characteristics which the methods in the paper.Large amount of simulation experiment has been made, results show that the proposed method can overcome above multiple environmental changes, has high matching accuracy and practical values. But because of the limit of time and individual level need more research and discussions. In code aspect need more optimized so that accelerate the matching speed.