Research On UAV Aerial Image Location

The positioning of UAV aerial images is a research hotspot in the field of computer vision,which has great application value in the fields of vision positioning and system navigation.However,due to the difference between remote sensing image and aerial image,image matching is difficult.For this problem,the thesis studies from three aspects: panoramic mosaic,geographic registration and Image location,The main work is as follows:Firstly,when SURF algorithm extracts features from high-resolution aerial images,the number of feature points is large and the distribution is dense,which leads to too many feature matching pairs and time-consuming.This thesis proposes to use the response strength of feature points to screen feature points,and to ensure the uniform distribution of feature points,distance constraint is applied to the reserved feature points.This method can effectively improve the speed of feature matching,and can effectively filter the feature points that fall in the open area and have unstable response.For image fusion,the stitching line search algorithm and multi-band fusion algorithm are combined to eliminate the ghost and improve the mosaic quality of the panorama.Then for the road extraction in geographic registration,there is a large noise in road extraction due to the interference of the ground objects around the road and the roof.In this paper,a comprehensive road extraction method based on connected region is proposed.Firstly,the image is preprocessed to filter out most of the background information,and then the connected region of the image is extracted.According to the characteristics of the road,only the connected region of the road is retained.Finally,the Hough transform is used to extract the road route.This method can effectively filter out most of the noise.Finally,for the realization of aerial image location,due to the feature matching directly can not meet the needs of real-time location,the thesis proposes a secondary matching strategy based on image retrieval.the feature coding algorithm is used to retrieve the sub image with the highest similarity to the aerial image,the feature matching algorithm based on mesh motion estimation is used to obtain more stable feature matching.The results show that this method is effective and has high positioning accuracy.