Study On Feature-Based Remote Sensing Image Registration And Change Detection

Remote sensing image change detection is the procedure of detecting and analyzing the change information between two or more remote sensing images,which obtain from same area on different times.With the development of remote sensing technology,remote sensing image change detection is widely used in forest environment detection,urban planning,natural disaster evaluation,military surveillance,and so on.Image registration is the general procedure before change detection.The result of image registration will have great influence on the result of change detection.Our research is focused on feature based remote sensing image registration and change detection.The mainly research is described as follow:(1)A feature matching based remote sensing image registration method is proposed.The random sample consensus(RANSAC)algorithm cannot receive the satisfying results when the accuracy is relatively low.The proposed method divides the data set into two parts through two threshold values.Sample set has high correct rate and consensus set has a large number of correct matches.An iterative method is put forward to increase the number of correct correspondences.The performance of the proposed method is validated on several datasets.(2)Remote sensing image change detection is studied,and convolutional neural network(CNN)based remote sensing change detection method is proposed.Convolutional neural network is a network structure which combined neural network and convolution operation.It can extract more natural features and has robustness.The proposed method receives the training samples through pre-classification,and these training samples are used to train the network.The final change detection result is obtained through the trained network.The results on different datasets show the accuracy and robust of the proposed method.