Research On Feature Point Matching Algorithm Which Based On Feature Point Density

In the empty triangle photogrammetry,the overlapping images on the route rely on the image connection points for the splicing operation.The extraction and matching of connection points is the main content of this thesis.Feature point extraction and matching is the core content of image processing and machine vision.It is widely used in target recognition,stereo imaging,dynamic tracking and other fields.In the field of photogrammetry,feature point matching technology also occupies an important position,and its effect directly affects measurement accuracy.In recent years,with the improvement of the performance of image acquisition equipment and the advancement of computer performance,it has laid a foundation for realizing the real-time processing of image feature point algorithms.This paper analyzes and summarizes the classical feature point extraction and matching algorithms,and introduces several image feature point extraction algorithms and their advantages and disadvantages.A diffusion recursive algorithm is proposed to find high-density feature point regions.Corresponding relationships are obtained by matching these high-density regions to calibrate left and right images to achieve scale invariance requirements.The main contents of the study are as follows:1.Selecting the current mainstream feature point extraction operators: SIFT,Harris,Fast,Moravec,etc.,summarizing the advantages and disadvantages of each operator,combined with the requirements of digital photogrammetric image matching,select Harris as the feature point extraction operator.2.The region extraction algorithm based on feature point density is studied.The image is meshed according to the size,and the high-density mesh unit is determined according to the number of feature points.The diffusion recursive algorithm is studied to extract the high-density cells connected together to form a high-density region.The high-density regions between the stereo images are matched,and the geometric position calibration is performed on the images according to the matching results.3.The algorithm of feature points in the autocorrelation screening reference image is studied,and the feature points with the most difference are found as the personality points,and the personality points are matched.4.According to the feature point correlation parameter,the least squares fitting of the quadratic function is performed to achieve the sub-pixel matching of the feature points.5.Prepare image matching program and conduct experimental analysis on research results.