Feature Analysis Of Image Control Point And Research On UAV Image Matching In Open-pit Mine

UAV photogrammetry is the main way to obtain image data for current earth observation.The acquired images have the characteristics of high resolution,short cycle and strong timeliness.In the production management of open-pit coal mines,it is an important task to carry out periodic UAV photogrammetry for open-pit mines and obtain DOM and DEM data for calculation of open-pit mine engineering quantities.In the process of UAV photogrammetry in open-pit coal mines,there are the following two problems that need to be improved: First,openpit coal mines need to periodically and repeatedly obtain images of the mining area,and the acquisition of photogrammetric ground control point data still relies on manual field collection.Second,the currently studied image matching algorithms have different applicability to UAV images in different regions.Open-pit coal mines contain a variety of ground object types,and the surface morphology and corresponding image characteristics of different regions are different.Commonly used image matching algorithms are used in image matching in open-pit coal mines,but the matching accuracy and matching accuracy are not high,which affects the quality of image matching and stitching.In view of the above problems,this paper mainly conducts two researches: First,on the basis of traditional ground control points,the author studies the selection and use of image control points in open-pit coal mines,and uses image control points instead of traditional field control points to provide Image control information.Second,in view of the problems existing in image matching in open-pit coal mines,the matching results of common image matching algorithms in open-pit coal mines are compared,and the algorithm with better matching results is selected to improve and optimize,so as to improve the quality of image matching.The specific research contents and results of this paper are as follows:On the basis of summarizing the types and characteristics of survey control points,four types of image control points in opencast coal mines are proposed,which are geodetic control points,ground mark control points,image geometric feature points and image feature points.The characteristics of four image control points are analyzed.The images of ground marker control points and geometric feature points of image features are mainly based on geometric structure features,while geodetic control points are mainly based on texture features.According to the characteristics of different types of image control points,the matching method between the control point image and the target image is analyzed,and the matching experiment between the control point image and the target image is carried out by using the gray-based image matching method and the feature-based image matching method respectively.Five sets of images from different functional areas of open-pit coal mines were selected as experimental data,and three algorithms of SIFT,SURF and KAZE were compared and analyzed.The experimental results of UAV image matching in open-pit coal mines show that KAZE has the largest number of correct matching points among the three algorithms,and can also match feature points in areas with weak edge information,and the matching accuracy is slightly higher than that of the SIFT algorithm.The matching time is similar to the SIFT algorithm.Among the three algorithms,the KAZE algorithm is more suitable for UAV image matching in open-pit coal mines.However,However,the KAZE algorithm takes a long time,and the matching accuracy and matching accuracy need to be improved.Aiming at the problems existing in the KAZE algorithm,an improved algorithm based on KAZE is proposed.In the feature extraction stage,the feature points of the KAZE algorithm are used.In the feature description stage,the binary descriptor FREAK is used instead of the floating-point descriptor M-SURF for feature description.In the feature point matching stage,the relative distance is used to further screen the initial matching points,and finally the Progressive sample consensus(PROSAC)algorithm is selected to replace the Random sample consensus(RANSAC)algorithm to eliminate false matches.Calculate the homography matrix to complete the image matching.The UAV image of the open-pit coal mine is selected,and the method in this paper is compared with the KAZE algorithm,The experimental results show that the improved algorithm based on KAZE in this paper improves the efficiency of the algorithm,the distribution of matching points is more uniform,and the matching accuracy and matching accuracy rate are significantly improved.