| This paper studied the problem of constructing mosaics from image sequences acquired in the air. It is the hot point of research, referring to image processing,computer vision,photogrammetry and so on.Because of many reasons ,the camera is not stable when it screens the scene of ground in the air.Its position and stance are easily effected. As the camera would lean and rotate,The images caught by the camera would have serious geometrical distortion. So we must transform the images firstly and the reaserach on camera-calibration is necessary.It is necessary to design some appropriate methods to solve the problems we faced. Firstly,The camera instrinsic will not change any more when they are fixed,for this reason and the precison. The calibration based on three-dimension objects was adopted. Besides, The position of the camera is ruleless in the air.So we studied the geometry constraint between image sequences and tried to get the extrinsic of each two images. After that,The algorithms for geometric rectification of oblique images were presented,which based on homography theory.Images acquired by the same camera in different orientation but with the same camera center is related by a Homography matrix.The homography could be calculated and used to complete the rectification.All the rectified images are ready for matching.According to the characters of image sequences, the matching algorithm based on plane-induced Homography was selected,which used the corresponding points to determine the Homography between adjacent images.Then, we used the way of conversing mapping and bilinear interpolation to transform the coordinates.So all images can be placed into the same coordinates and pieced together as a bigger image.For the precision,we used some algorithms to improve the quality of results.By designing the experiments based on the image sequences.we analyze the algorithms in this paper. The results indicated that the algorithms were feasibility. |
PDF Full Text Request