Classification of remote sensing image data fusion considering spatial information

Classification is an important application in remote sensing image process, but classification accuracy about non-normal distribution of samples is lower when using traditional methods of probability and statistical. Evidence theory can combine certain and uncertain information of multi-source remote sensing images to achieve effective identification of the images. Taking the results which go through training of neural network as evidence can combine the neural network with the evidence theory, and then integrate their advantages to get better classification results. In the paper, we proposed the classification of remote sensing image decision-level data fusion considering spatial information, and took the panchromatic image with plentiful spatial information into classification decision to reduce uncertainty, and to improve the classification accuracy.