Collaborative classification method fusing complete polarimetric SAR and hyperspectral remote sensing

The invention discloses a collaborative classification method fusing complete polarimetric SAR and hyperspectral remote sensing. The method comprises: performing polarimetric decomposition and texture feature calculation on a synthetic aperture radar (SAR) in the same region, and re-sampling a hyperspectral image and an SAR image to the same resolution; converting the sample vector file into raster data, performing label extraction and taking the raster data as a training sample of a classifier; calling machine learning methods such as a maximum likelihood method (ML), a mahalanobis distance method (MD) and a support vector machine (SVM) to carry out classifier model training on the image training data set; stretching the resampled images into different one-dimensional feature vectors, and then combining the one-dimensional feature vectors into a prediction data set; and calling a previously trained model to predict a label value. The method has the advantages that the method can be directly applied to comple