Research On Matching Algorithms For UAV Image Of Karst Plateau Mountain Area

The terrain of Karst Plateau Mountain Area is complex,deep gully,and the traditional methods of obtaining geographic information are difficult and costly.Using UAV technology to obtain the geographic information in this region has the advantages of fast acquisition speed,high efficiency,low cost,and high spatial and temporal resolution of acquired images.Using UAV aerial survey technology to obtain geographic information in this region has become an irresiistible trend.The images obtained by UAV in karst plateau mountain area have such phenomena as large deformation,weak texture and local shadow,which make it difficult to match UAV images in this area.In order to improve the matching accuracy and efficiency of UAV image in Karst Plateau Mountain Area,this paper carries out matching research on the orthography and tilt images of UAV images in Karst plateau mountain area.The main research contents and results of this article are summarized as follows:(1)Aiming at the problem that poor quality of matching points in orthophoto matching of UAV in Karst Plateau Mountain Area,this paper constructs a method for accurate orthophoto matching of UAV in karst plateau mountainous area with RMSE PROSAC algorithm.The algorithm firstly refines the coarse matching points by the PROSAC algorithm to obtain the initial "inner points",and then all the initial "interior points" are filtered again by RMSE,and the "interior points" after secondary screening are taken as the fine matching points.The experimental results show that the method constructed in this paper,which adopts the PROSAC algorithm combined with RMSE to carry out accurate orthoprojective image matching of UAV in karst plateau mountain area,can obtain more matching point pairs with better quality than the PROSAC algorithm.(2)Aiming at the problem that KAZE algorithm in UAV orthophoto Image Matching of Karst plateau mountainous area,which takes a long time and has fewer excellent feature points,this paper constructs an orthophoto matching algorithm for UAV in Karst Plateau Mountain Area based on the KAZE-DAISY algorithm.The algorithm firstly uses the KAZE algorithm to extract image feature points,then uses the DAISY descriptor to describe the obtained feature points,and uses the FLANN algorithm and the ratio purification method to perform rough image matching.Finally,the PROSAC algorithm combined with RMSE is used for fine image matching.The experimental results show that compared with the KAZE algorithm,the algorithm in this paper can obtain more excellent feature points in the case of shorter time-consuming matching in UAV orthophoto image matching in karst plateau mountainous areas.And the algorithm presented in this paper performs better than KAZE algorithm,SIFT algorithm and BRISK algorithm in UAV orthophoto image matching in karst plateau mountainous areas by comprehensively considering the indexes such as total matching number,correct matching number,matching accuracy and total matching time.(3)Aiming at the problem that the traditional algorithm can't match the UAV's tilt image well or even can't complete the matching,this paper constructs a matching algorithm of UAV's tilt image based on ABRISKE algorithm.The algorithm firstly uses a completely affine invariant frame to simulate the perspective of the original image,then uses the BRISKE algorithm to extract and describe the feature points of the simulated image,and obtains the feature points and feature descriptions of the original image,and uses the FLANN algorithm as well as the ratio purification method to carry out coarse image matching,finally use PROSAC algorithm based on homography matrix to perform fine image matching.The experimental results show that the algorithm based on abriske can achieve more excellent feature points and better complete the matching of UAV tilt image in Karst plateau mountain area compared with the traditional algorithms such as crisp algorithm,SIFT algorithm and surf algorithm.