Research On Cooperative Positioning Technology Using Multi-scale Remote Sensing Images

With the massive acquisition of multi-source and multi-scale remote sensing images, inorder to reduce the workload of fieldwork, cooperative positioning technology using multi-scaleremote sensing images become a research focus in recent years and have a significant foregroundof application. Considering the characteristics of cooperative positioning based on multi-scaleremote sensing images, this article gives more in-depth study and exploration on the multi-scaleimage matching, high-precision control information obtaining, stereo image positioning, systemerror compensation and so on. The study contents and innovations are listed as follows:1. Multi-scale space theory is researched. SIFT algorithm and SURF algorithm areparticularly analyzed. Experimental results show that, compared with SURF, SIFT gives moreright match points and better match point accuracy. Consequently, SIFT is the preference methodin multi-scale image matching. According to the degree of the scale difference, the appropriatematching threshold can be selected, which is important.2. A transfer method of control information in multi-scale image is presented, based on thecombination of scale invariance features and gray information. In this paper, SIFT is chosen asthe initial match to rectify the target image. After that, with the help of gray information, thereference image and the corrected image are matched using the correlation coefficient and leastsquares matching method and then error matching points are eliminated, so more precise andreliable matching points are gotten. Experiments show that in the region where the match pointsare sparse, the multi-scale image control information transfer methods based on the combinationof scale invariance features and gray information can get better accuracy control points thandirectly using homography method.3. Remote sensing image sensor models are described and analyzed. Stereo positioningmethod based on rational function model and system error compensation methods of satellitestereo image positioning are summarized one by one. Experiments results show that thepositioning accuracy has been greatly improved with control points joined in the system errorcompensation. The control information transfer method presented in this paper shows bettereffect of system error compensation in object-space.4. Two methods that object image rapidly rectify are presented based on orthophotos asreference, using this two methods to produce orthophotos have the characteristics that easily toobtain raw control data and quickly to fulfill work, these will play a significant role in rapidresponse.