A Heterogeneous Image-Matching Algorithm Based on a Multi-Level Screening Strategy

In order to solve the problems of low confidence and poor correlation of feature point pairs between images due to different imaging principles and significant modal differences in the heterogenic image alignment task. This paper proposes a heterologous image-matching method based on a multi-level screening strategy based on the binocular camera's imaging principle and relevant a priori knowledge of image alignment. Firstly, we remove the apparent error points in the feature point pairs according to the feature points' relative position and main direction through the feature point pair screening strategy based on the search area and the feature point pair screening strategy based on the slope consistency. Then, the final screening optimization of feature points is carried out by improving the RANSAC algorithm to improve the quality of image alignment further. Finally, to verify the performance of the alignment algorithm proposed in this paper, the alignment results of this paper's method on several heterogeneous image datasets are compared. The experimental results show that compared with other algorithms, the alignment algorithm proposed in this paper has significant advantages in accuracy, stability, and matching efficiency performance indexes.