Multispectral And Panchromatic Images Registration And Change Detection Technology Research

The image registration and classification have become the hot topic for image recognition and processing, which have been widely studied by experts at home and abroad. They are not only the key steps for some intelligent recognition and image processing, but also the main parts of many industrialized application. For example, they are widely used in the field of computer vision, multi-spectral and panchromatic image fusion, medical imaging analysis, etc.This paper mainly discusses on multi-spectral and panchromatic image registration and change detection of various technical indicators in detail according to the experimental results, which has also analyzed the process at the same time. This paper inquired the algorithm based on SIFT feature extraction method and registration in a comprehensive and detailed way. Firstly, it transformed the two images into the same coordinate system through projection transformation, and then the overlap region in the registration image which had the same scale, same angle with the reference image was extracted. Secondly, in the process of extracting the feature points, this paper made some improvements in the SIFT method. Then the Delaunay triangulation was used to retain the matching points without deleting the original information. Finally, a more precise image transformation matrix was given to obtain the optimal results, initial value. The method proposed in this paper improved the registration accuracy and speed. The evaluation of this method was divided into two parts:image fusion as the subjective one to evaluate the accuracy of the registration area; the root mean square error and correlation coefficient are objectively evaluate the registration areas.The change detection technique of multi-spectral and panchromatic images is another important branch of digital image processing. The technology mainly studied the different time points of the same object, different scene and analyzed the reasons for the changes. Due to the changing detection method of multi-spectral and panchromatic images. the image classification technology was used. The paper makes classification for multi-spectral and panchromatic images by using the BP neural network, and analyzes the experiment results. The paper improved the previous the method of difference method for change, making change detection by difference method after the panchromatic and multi-spectral images was classified, and then the results were analyzed.