Research On Image Registration With UAV Recon Video In Plateau Area

In recent years, the UAV aerial technology develops rapidly and people utilize the UAV mounted with a camera for low altitude aerial photography to acquire high resolution remote sensing image data of the target region. As an effective technology to supplement traditional manned aircraft aerial and satellite remote sensing, the UAV aerial technology has the advantages of automation, low cost and high efficiency and plays a key role in military terrain reconnaissance, the debris flow hazard assessment, and animal and plant ecological survey and other practical applications. Low altitude aerial photography with a UAV in the plateau area to obtain video data of the target region and mosaic of UAV video image sequences of the plateau area to the target region's panoramic image has great practical value.a single image's vision field is small, covering only a part of the region and small phase and amplitude and large amount is the common characteristics of UAV recon video image. Including these common characteristics, UAV recon video image of the plateau area has its own characteristics:(1)image texture is singe, mainly desert and frozen soil;(2)UAV is small and easy to tilt and jitter, resulting that the stability and wind resistance ability is bad and images have serious geometric distortion. These characteristics are peculiarly prone to make the situation of mismatch happen during feature extraction and feature matching. Besides, due to the complex environment of the plateau area, we cannot realize Image geometric distortion correction by setting ground control points. All the above have brought many challenges for UAV recon video image panorama stitching of the plateau area.The feature point of UAV aerial video data of the plateau area is not obvious, leading to the problem of mismatch when adopting the image mosaic algorithm based on feature. For this, this paper researches on utilization of inertial navigation system(INS) data information to improve the image mosaic algorithm based on feature and finally stitches Multiple overlapped UAV aerial video images of the plateau area into a smooth seamless, high resolution panorama image. The main work and innovations are as follows:1. An introduction of the acquisition situation and characteristics of UAV recon video data in plateau area is given. Based on this, this paper further analyzes and the reason why the traditional stitching algorith doesn't work and summarizes the key problems that need to be solved for mosaic of UAV recon video images of the plateau area. Combined with The working principle of the UAV inertial navigation system, it has discussed The main error sources Causing UAV recon video image geometric distortion2. A UAV recon video image Geometric correction algorithm based on inertial navigation error compensation is proposed. In the light of the problem of low precision of UAV inertial data information, this paper has established an error compensation model based on Kalman filtering, which is used to correct location parameters and flight attitude parameters of UAV recorded by INS; In the light of the problem that the plateau region is unable to establish ground control points, this paper has established an UAV recon video image geometric correction model based on the change of flight attitude.3. An UAV recon video image registration algorithm based on multiple information fusion is proposed. This algorithm makes the registration algorithm based on feature match and INS data information combined and complemented with each other and achieves effective stitching of UAV recon video images of the plateau area eventually.Through experiments on multi segment real UAV aerial video data of the plateau region, algorithms proposed in this paper are proved to be feasible and effective and can be well applied in panorama stitching for UAV recon video images of the plateau area.