Research On Key Technologies Of UAV Remote Sensing Image Mosaic

UAV remote sensing is a new remote sensing method, which h as the characteristics of fast, flexible, low cost, high image resolution and so on. It has developed rapidly in recent years. UAV remote sensing has been widely used in many areas, including agriculture, forestry, electricity, land resources, city planning and so on. Especially in agriculture, unmanned aerial vehicle(UAV) is widely used in crop monitoring, plant diseases and insect pest monitoring, crop nutrient analysis, crop growth monitoring and other aspects which can help agricultural production pers onnel acquired the crop growth informationtimely and accurately. Also, it can provide scientific basis for disaster reduction and emergency command, which has achieved a more ideal application effect, and has broad application prospects.However, the UAV remote sensing platform in the process of aerial photography by flight height, camera focal length restrictions and so on, acquired in a single image coverage range is small, which can’t cover the desired area. Therefore, in order to get the entire target a rea information, that need to get more than a remote sensing image fusion splicing into a panoramic image. UAV remote sensing image mosaic is the basis of remote sensing interpretation and analysis. The quality and effect of the splicing directly determine the accuracy of interpretation and analysis. Therefore, to carry out the UAV remote sensing image splicing technology research has important practical significance.In this paper, the domestic and foreign remote sensing image stitching technology was in-depth research and study, comparison and analysis of the characteristics of the existing remote sensing image splicing technology. Based on remote sensing image stitching process, focus on the development of the UAV remote sensing image preprocessing techno logy, remote sensing image registration techniques and remote sensing image fusion technology research. The main research work is as follows:(1) Remote sensing image preprocessing. Remote sensing image preprocessing is a very important step to remote sensing image mosaic, which directly determines the precision of image stitching. UAV in the interval between the two aerial, due to the light intensity and the incident angle changes, the two adjacent images will have a larger color difference, image brightne ss, saturation, and so on. Therefore, it is necessary to radiometric correction of the image to eliminate the influence of the image color difference on the stitching. Since the UAV in the process of aerial photography, two adjacent images shooting time is very short, and the difference is relatively small influenced by the atmosphere and illumination. Therefore, this paper uses the histogram matching method for relative radiometric correction, by adjusting the reference histogram of the image and makes the images histogram matching. According to the correction results, the corrected image is basically consistent with the brightness and hue of the image to be stitched, and the smooth transition of the two images was well achieved. In addition, as UAV has the characteristic of small volume, light weight, air current is large and the stability and wind resistance ability is poor, flight attitude inclined, jitter phenomenon is difficult to avoid. All of these will access to remote sensing images produce direct effect, lead to the distortion of the image. In this paper, the geometric correction of distorted images is carried out by means of image point coordinate transformation and image resampling, in order to eliminate the influence of geometric distortion on th e image, and to meet the need of image registration.Geometric correction has the characteristics of large computation and long time. In order to improve the processing speed of geometric correction, this paper proposes a parallel geometric correction algorithm based on distributed processing for UAV remote sensing image. The algorithm makes the data needed by the processor to be stored in the local, avoids the communication between the processors, and solves the problem of local data, and improves the efficiency of parallel processing. In the cluster system, the parallel geometric correction and the serial geometrical calibration algorithm are tested, compared and analyzed. The experimental results show that the parallel geometrical calibration algorithm ha s good parallel performance and greatly improves the processing speed of geometric correction of remote sensing image.(2) Research on remote sensing image registration technology. Image registration is the core of the image mosaic technology system. The precision of image registration directly determines the quality of image mosaic. Due to the SIFT algorithm for rotation, translation, scaling, brightness change to maintain invariance, the perspective changes, affine transformation, noise is also to maintain a certain degree of stability. Therefore, the algorithm is very suitable for UAV remote sensing image registration. In this paper, the feature points extraction based on SIFT algorithm.In the process of feature points matching, because of the self-similarity of adjacent images, one to many or many to one error matching must be eliminated. In this paper, we proposed a method to eliminate the error matching feature points by using the epipolar geometry constraint and the homography constraint. Firstly, a part of the error feature points were eliminated by reducing the search scope of the epipolar geometry constraint. Then obtained the correct matching points to establish the homography matrix transformation model based on further eliminate error characteristics of image matching, to improve the accuracy of matching. Use of RANSAC algorithm to remove the outer corner, purification of matching feature points, that the precise matching of image feature points was realized.(3)Research on remote sensing image fusion technology. The purpose of image fusion is to merge the registration image based on a certain transformation model into an image, and the two images should be seamless splicing, and no trace of stitching. Remote sensing image acquired by UAV, due to the impact of illumination change and sensor incident angle change, between the adjacent images will differences in brightness and color. If the image directly for splicing, in the overlapping area of the two images will form the obvious mosaic trace, that the emergence of the phenomenon of "ghosting".To solve this problem, this paper proposed a method using optimal seam-line to eliminate "ghosting". Firstly, the optimal seam-line search based on dynamic programming was adopted. But using this search strategy, if there were some error points in a seam-line will lead to "ghosting" can’t completely eliminate, the seam-line obtained is a local optimal seam-line, that can’t get the global optimal seam-line. To this end, the optimal seam-line search based on graph cut was proposed. In seam-line search, maximum flow of directed graph based on maximum flow minimum cut theorem, was optimal stitching line. Two images of the location of the mosaic was more accurate, to get a better solution on both sides of the joint dislocation and the "ghosting" phenomenon, that realized the real seamless splicing. At the same time, the image can be smoothed out by using the poisson fusion method because of the difference between the brightness and chroma of the mosaic image.