Socier surface moraine remote sensing identification method based on feature optimization random forest

The invention discloses a glacier surface moraine remote sensing identification method based on a feature optimization random forest, and the method mainly comprises the following steps: carrying out the radiometric calibration, atmospheric correction and resampling of a sentinel 2 remote sensing image, and splitting the synthetic spectral band of the image into a single-band image; constructing four types of feature vectors of spectral information, remote sensing indexes, image textures and topographic features, and carrying out importance sorting on all feature variables; sequentially superposing feature vectors on importance sorting results to train an RF model, and finding out the number N of the feature vectors when the moraine recognition precision reaches the highest; and with obtaining of the highest surface moraine identification precision as a criterion, preferably selecting the first N feature vectors to form an optimal feature combination, training an RF model and carrying out image classification