Research On Key Technologies Of Automatic DSM Extraction From Multi-view Oblique Images

Multi-view oblique camera is valued for its good performance in the 3D reconstruction due to the ability of obtaining high overlap, multi-view images. However, the oblique images contain severe geometric distortion and radiometric distortion, while the occlusion phenomenon is also more serious, which makes the DSM(Digital Surface Model) extraction more difficult. Therefore, methods on the automatic DSM extraction from oblique images is discussed in this paper. Based on the analysis of the oblique images’ characteristics, technologies based on the Multi-view oblique images such as tie point extraction, bundle adjustment and dense matching have been studied, and automatic DSM extraction process has been given. The major work and achievements are given as follows:1. Based on the analysis of AMC580 camera parameters, the preprocessing of images, such as image de-noising, image enhancement, camera distortion correction and epipolar images generation are realized. The image preprocessing can provided high-precision epipolar images for the image matching in the following part.2. Aimed at the serious distortion on the oblique images, in terms of the points’ coplanar relationship, a large oblique stereo images matching algorithm based on projective relation is proposed combined with SIFT and Shi-Tomasi features. The experiments show that the algorithm is adaptive to the large oblique stereo images matching.3. Attend to the characteristics of the images obtained by the multi-view oblique camera, a down-view based scheme for tie points extraction algorithm is proposed. A mount of high-precision tie points is extracted by the scheme.4. The basic principle of bundle adjustment is elaborated. Based on the analysis of the settlement method on nonlinear model, LM algorithm is introduced to solve bundle adjustment equation, and the tie point obtained by the strategy proposed in this paper is used to solve bundle adjustment, which improves the precision of exterior orientation elements effectively. Experimental results show that LM algorithm can obtain better position accuracy than the traditional algorithm.5. Semi-global matching based on BT algorithm is introduced to extract dense matching point from the same-view images, and the process of automatically extracting DSM from multi-view oblique images is analyzed. The adaptive weighted fusion strategy based on down-view DSM is proposed. Multi-view oblique images DSM is extracted Effectively.