Research On Technologies Of Reconnaissance Image Mosaic And Map-matching

Image mosaic and map-matching is an important study field in the digital image processing. In order to get the large field image and realize the generation of panoramic image, image mosaic technology formats for a large field image by matching, aligning and then fusing the double or multiple partly overlapped images, widely used in the fields of remote sensing, medical science, industry and agriculture etc.. As a foundation of other workings, the studying of map-matching has a vital significance in the processing of remote sensing.The main content of this dissertation is the study of image mosaic and map-matching technologies of the Unmanned Aerial Vehicle reconnaissance images. Meanwhile, the aligning and mosaic of the UAV image is emphasis discussed in this studying. Initially, the dissertation elaborates the processing of the image mosaic and the key skill of the realization. Then, the theoretical basis and key skills of map-matching are recommended. Last, in order to get a pleasurable visual effect, it makes image registration and stitching according to the features of the UAV detection images.In the process of stitching of reconnaissance images, initially, mainly in the aspects of features from the extracted feature points, the detected speed and noise immunity, some classical detector algorithms of feature points were analyzed and compared explicitly by experiments. Then, according to the characteristics of reconnaissance images, SIFT (Scale Invariant Feature) corner detector algorithm is finally adopted to collect key points. In order to improve the image matching accurate, NCC (Normalized Cross Correlation) combined with Mahalanobis Distance are also used to treat the key points detected by SIFT algorithm. As a result, the forged matching points are decreased and new match points are acquired. Last, the remote detect image is stitched and fused by utilizing the new matching points. Compared with the existing image mosaic methods, this method can get more accurate matching matrix with better effectiveness and robustness.