A Change Detection Algorithm for High Resolution Remote Sensing Images based on Data Analysis

With the development of remote sensing technology to high resolution, high resolution remote sensing images are widely used in change detection. This paper uses Convolutional Neural Network (CNN) as its core model. The first step is data preprocessing, including data cutting, re-projection and normalization. Then the output of post-processing and change model is refined by threshold processing and morphological operation. Experiments show that the maximum detection time of CNN is only 400 milliseconds, showing higher speed performance. In terms of detection accuracy, the best CNN accuracy rate is 96.6%, and the worst CNN accuracy rate is 72.3%. The highest accuracy rates of RNN and SVM are 69.4% and 67.3%, respectively. As far as the reliability of detection is concerned, CNN has the best reliability, which is 96.9%, RNN is 79% and SVM is 72.2%. These results show that CNN can achieve better performance in typical high-resolution remote sensing image change detection tasks.