Multi-perspective Templates Correction And Stitching Method Based On INS

High-resolution,large field of vision of the target templates are the prerequisite for image precision-guided. Image stitching technology is a very good solution to the contradiction between resolution and field of vision of target templates. However, large perspective difference between aerial images is an important factor in the effect of image stitching, and the aerial image itself is not easy to obtain increase the difficulty of experimental demonstration.In this paper, an overview of image stitching technology is giving firstly, and then it details the study of the status quo at home and abroad.In order to eliminate the effects from perspective,this article take aerial image carrier INS into image correction, get weak perspective projection images as corrected image according to the perspective projection image. Aerial imagery is difficult to obtain, to solve the problem, this paper designed a simple and feasible method, capture images on the ground, similar to the multi-view aerial images as a correction trial.In the image registration section, this article details the SIFT feature based image registration algorithm first.As this algorithm has some disadvantages like high computational cost and poor real-time performance. To solve this problem, an improved SIFT algorithm based on self-adaptive threshold FAST was proposed. The method combines the FAST corner detection algorithm into the feature detection process of SIFT. Through a preliminary selection of the candidates, the feature detection process of SIFT is highly speeded up, and the total time needed for matching is much less than before. As it is difficult for the FAST algorithm to determine the threshold artificially, and it is easily disturbed by noises, an improved FAST algorithm was presented, in which the threshold is determined automatically according to the image gray distribution without manual intervention.In the image fusion section, paper first describes the common methods and theoretical framework of image fusion, and then choose the fusion method based on multi-resolution analysis. According to that gray level co-occurrence matrix (GLCM) can reflect the characteristics of grain roughness and direction accurately, paper propose a fusion strategy of multi-resolution image which combined with GLCM: First, calculate the GLCM of the low and high frequency part of the multi-resolution image, then, the energy of the low-frequency GLCM and the contrast of the high-frequency GLCM corresponding as activity measure for image fusion to selecte the low and high frequency coefficients. Finally, the paper using a linear transition to eliminate stitching line, then achieve seamless stitching image.