Integration of Radiation-Variation Insensitive Feature Transform and Fast Sample Consensus Algorithm for Automatic Multimodal Remote Sensing Image Registration

In view of the limited feature points extracted by traditional algorithms in heterogeneous remote sensing image matching, resulting in a small number of successfully matched point pairs and the need for improved matching accuracy, a method combining radiation-variation insensitive feature transform (RIFT) and fast sample consensus (FSC) algorithms is proposed. Firstly, the RIFT algorithm is used to extract image features to form preliminary matched point pairs, which is robust against non-linear radiation distortion and suitable for feature extraction from different modal images. Then, the FSC algorithm is employed to geometrically filter the preliminary matches, removing abnormal matches to enhance the accuracy of the matching results. Experimental results show that, compared with traditional algorithms, the method proposed in this article improves the average matching accuracy by approximately 16.9 %. The conclusion indicates that the matching method integrating RIFT and FSC algorithms can significantly enhance the matching effect of multi-modal remote sensing images, catering to the highprecision matching needs of complex remote sensing data.