Research On Fast Detection And Matching Method Of Image Feature Points

Feature point detection and matching techniques are widely used in many research fields such as image stitching,3D reconstruction,and robot navigation.Therefore,the research work on feature point detection methods,feature description and matching strategies is as follows:(1)In order to improve the accuracy of the detection results of Shi-Tomasi algorithm and speed up the calculation,a Shi-Tomasi feature point detection algorithm based on gridded model is proposed.This method makes the contrast more visible by using the lightness channel of the HSV color space.The Laplacian operator is used to extract the image edge information features,and the invalid feature value calculation is filtered out.Then the image is meshed and segmented,the local threshold of each image is calculated by the Stomasie algorithm,the noise interference is reduced,and the OpenMP parallelization technology is combined to accelerate the processing.Finally,the feature points of the block detect.ion are integrated to complete the feature point detection.In the experiment,the method is compared with the world's three algorithms,and the results show that the method has higher detection accuracy and speed.(2)In order to strengthen the fine matching constraints of GMS algorithm and improve the accuracy of registration results,a GMS matching algorithm based on feature consistency is proposed.The method introduces feature consistency constraints,and optimizes the GMS matching model by using the correct matching pair with the same motion trend and one-to-one relationship.On the basis of this,the rotation invariant rBrief feature descriptor and the improved GMS model are combined to obtain the matching pair with one-to-one relationship,which reduces the mismatch rate in the fine matching phase and completes the feature matching.The experimental results show that compared with GMS and other matching algorithms,the registration ac.curacy of feature points is improved.(3)In order to realize image stitching based on feature points,a stitching method based on the combination of block detection and motion constraint is proposed.The block detection,feature constraints and pixel point fusion methods are merged into the splicing framework.First,the image feature points are detected by the Shi-Tomasi detection algorithm based on the block image.Then,the mismatching pair in the rough matching result is filtered out by the motion constraint condition of the feature point.Finally,the image transformation relationship is estimated by using the fine matching pair,and the pixel point distance fusion method is used for splicing.This method speeds up the stitching and improves the accuracy.