Study On Point Feature Matching Technology Facing The Data Preprocessing Of DMZ Camera

With the development of photogrammerty technology, high resolution format CCD digital aerial camera has become one of the main methods of access to the aerial remote sensing data, and the image preprocessing technology has also become a research hotspot of photogrammerty workers. For multi format digital aerial camera, image matching is an important part of the preprocessing technology of camera images. This paper mainly focuses on the image matching technology in the preprocessing of a new domestic multi format digital aerial camera(DMZ camera) images. According to the characteristics of DMZ camera images, this paper proposes the matching algorithm for DMZ camera panchromatic images, and the matching algorithm between panchromatic images and multi spectral images, and verify the effectiveness and applicability of those algorithms by experiment. The main works of this paper include:1. Review the development of digital aerial camera, describes the background of the multi format digital aerial camera born and the necessity of develop the domestic multi format digital aerial camera system. Analyze the performance and the image preprocessing technology of several domestic and international representative multi format aerial cameras. Introduces a new domestic multi format digital aerial camera DMZ camera system. Point out the importance of the image matching technology in the data preprocessing.2. Introduce the basic concepts of image matching and several commonly used similarity measures. Analyze the principle of the image matching algorithms based on the image feature and gray, and point out the advantage of feature matching algorithm. Focus on the image point feature matching algorithm, lists several classic point feature matching algorithms from the early 70’s. Analyze the basic principle of Harris algorithm, SIFT algorithm, BRIEF algorithm and ORB algorithm. The characteristics and applicability of several commonly used image geometric transformation models are also analyzed.3. According to the characteristics of DMZ camera panchromatic images, apply the SIFT algorithm and ORB algorithm to the matching of panchromatic images. Then propose a matching algorithm more suitable for the DMZ camera panchromatic images. This algorithm is based on Harris algorithm and the normalized cross correlation coefficient, use the matching strategies like image block and local search and the culling algorithm based on the offset of image point coordinate, achieve high precision positioning through the least squares matching, and optimize the matching results by free network adjustment.4. According to the characteristics of DMZ camera multi spectral images, apply the SIFT algorithm to the matching between panchromatic images and multi spectral images, and test to analyze the feasibility and limitations. Propose a matching algorithm more suitable for the DMZ multi spectral images by combining with matching strategies such as coarse to fine, block-match and adaptive threshold.5. Design the image feature matching test software according to methods proposed in this paper. Test the images from Han Zhong area, the results include two parts: the panchromatic image matching results show that the matching accurate is 0.2~0.3 pixel, and distribution of match points is good, and the algorithm has good stability, the panchromatic and multi spectral image matching results show that the algorithm proposed in this paper is 3 times faster than the original SIFT algorithm, also has better distribution of match points, and the matching accurate can achieve 0.4 pixel. All above can verify the effectiveness and applicability of the proposed algorithm.