Uav Remote Sensing Image Matching Technology Research

Unmanned Aerial Vehicle (UAV) has advantages of efficiency, flexibility and flying ability, probing high-risk areas, low cost, and so on. so UAV-platform remote sensing imaging technology has been rapid developed and has been applied widely in the fields of municipal planning, non-human disaster supervision and military reconnaissance, and so on. At the same time, the UAV platform for remote sensing image pixel numerous, irregular, and its practical application considerable difficulty. In view of this situation, there must be used in the stitching algorithm.First, the present research on image mosaic technology with its application domain is introduced in detail, pointing out that the image mosaic technology based on feature-point is the main trend in recent years; Then studying the technology of geometric correction with regard to UAV remote sensing image. Second, by commenting on a few frequently-used techniques about SUSAN、Harris and SIFT based on the feature points, this thesis puts forward a new way combining the algorithm of SIFT and Harris; In the stage of image registration, the thesis combines k-d tree and Nearest-Neighbor with Distance Ratio to match features rough, then using RANSAC to purify the matching feature points. Finally the image mosaic is done with the improved SIFT and fading-in-and-out algorithm.The major contributions in this thesis can be summarized as follows:the very large number of feature points extracted by the SIFT algorithm, so if we use the SIFT directly on image, which will affect the efficiency of the image registration. Therefore, Harris operator is introduced in the SIFT feature point extraction process to enhance the uniqueness of the feature points, which pre-removes instability feature points lying the edge of the image. The experimental results show that the method of SIFT with Harris operator reduced the number of feature points significantly. This not only increases the stability of the matching significantly, but matching efficiency has been greatly improved, then improving the stability and real-time of the image mosaic.