Abstract: Random forest is an emerging and highly flexible machine learning algorithm with good stability in prediction and classification, and the performance of the algorithm is better than many single predictors. In view of this, a remote forest image change detection algorithm for random forests is proposed. The entropy rate method is used to superpixel segmentation of remote sensing images to obtain optimal segmentation results. A remote forest image change detection model based on random forest is constructed. The extracted Gabor features and spectral features are used as model inputs for training and prediction, and the decision tree voting is used as the final change detection result. The experimental results show that the random forest change detection model constructed in this paper is significantly lower than other algorithms in the missed detection rate and false detection rate, and the overall correct rate is the highest, and the algorithm time is also significantly better than other algorithms.

Key words: remote sensing image, change detection, random forest, superpixel, Gabor feature