Accurate Keypoint Mapping Of Multi-spectral Images Based On Cascade Structure

With the rapid development of information technology,single spectral images can not meet people's needs for image information,multi spectral image is more and more applied in various fields.In this paper,a keypoint mapping method based on cascaded structure is proposed for multi spectral images,and the accuracy of the registration results is improved by introducing the resurrection mechanism in the cascade structure.The keypoint mapping algorithm is a kind of common registration algorithm.The keypoints are detected in the reference image and the test image.The common methods of detecting keypoints and calculating descriptors include SIFT,SURF,MSER and so on.In this paper we use the SIFT method with higher accuracy and faster computing speed.In this paper,the cascade structure is composed of four layers.The first layer of the cascade structure is bi direction mapping.This step is also the process of constructing the initial mappings.The second level analyzes the intersections of the line connecting two keypoints and the detected lines from images.The third level delete wrong mappings with Global Information.The last level adopts FSC to remove wrong mappings and build the transformational model.Taking into account the process of deleting the error matching process,it is possible to have correct mappings to be deleted,so the introduction of the resurrection mechanism in the cascade structure to further ensure the accuracy of the registration results.Finally,the keypoints mapping of multi-spectral images are realized by using the cascade structure with the resurrection mechanism.In the experimental,644 pairs of image are used to carry out the contrast experiment.Finally,the experimental data shows that the proposed algorithm has a significant improvement in the registration performance compared with the previous classical algorithms.