Line Matching For Line-array Satellite Remote Sensing Images Based On BLDCB Descriptor

According to the theory and method research of photogrammetry and remote sensing,it has become a new development trend to realize three-dimensional reconstruction of target features by using stereo mapping images with line-array CCD push-scan imaging.In the 3D reconstruction of line-array satellite remote sensing images,the reconstruction process based on linear features can obtain reconstruction results with high structural expressiveness,and the quality of the final product is directly determined by the results of line matching.This paper proposes a bisected line descriptor with changing bands for line array satellite remote sensing image to address the problems of inconsistent line endpoints,low textures and unstable gradients of line in edge region.Firstly,the algorithm uses the RFM forward and inverse solution models to construct the corresponding epipolarities of reference image and the search image to achieve the epipolar constraint.Then,the Gaussian Laplace algorithm is used to detect the target zero crossing points on both sides of the extension of the straight line to construct a new straight line,and the new straight line is used as the center to construct a new straight line region,the BLDCB descriptor is constructed to achieve line matching.Lastly,the geometrical relationship of corresponding points in the local region of the line is used to constrain the one-to-one corresponding line for line array satellite remote sensing images for line matching.The experimental results show the validity and robustness of this method.This paper mainly focuses on the following aspects of the study:(1)Constrain the matching search range effectively.Based on the photometric characteristics of the line array CCD satellite remote sensing image,the RFM-based epipolar extraction is realized,The epipolar double constraints are used as the initial matching constraint to obtain the candidate matching set corresponding to the target line,and complete the task for the initial line matching.(2)Constructing line descriptors.The scale invariance of Gaussian Laplace algorithm is used to achieve the detection of new straight line endpoints and construct a straight line support domain with the new line as the center,sub-region,and increasing number of line bands.The support domain is divided into left and right parts each generating linear descriptors for line matching.Among them,the generalized Gaussian weight function is applied to increase the information weight of the central straight line between rows of the support domain and reduce the edge trailing.(3)Checking corresponding lines.The linear support domain is constructed to divide the support domain into two parts,and the Euclid distance threshold constraint from the corresponding points to the straight line in the support domain is used to achieve a one-to-one check of the corresponding lines.(4)Experiments and analysis of results.Eight sets of remote sensing images of ZY-3 satellite were selected to design line matching experiments with different parameters,and implementing line matching experiments with different line descriptor algorithms including LBD and N-LPI.The reliability of the BLDCB descriptor in this paper and it’s adaptability to the line-array satellite remote sensing images are verified by comparing and analyzing the experimental results.