Hyperspectral Remote Sensing Image Classification Based On Deep Learning

Abstract

Abstract In recent years, with the development of science and technology, in order to further explore the world, we have explored and studied remote sensing technology. As a new discipline, teleology has been widely studied and applied in the fields of spectroscopy, informatics, geography, environmental science and urban construction, and has become one of the most active fields of science and technology. However, both classifier and feature extraction are in the shallow level. How to extract deep features and make them more abstract and easy to classify is a hot issue in machine learning field. Therefore, through the study of different classifiers and various special cases, we find that the post-processing method proposed in this paper can alleviate this situation to a certain extent. When the number of hidden layers is 2, the overall classification accuracy of DBN model is higher. Experiments show that this method is better than the traditional deep learning method.