Remote Sensing Image Recognition Using Deep Belief Network

How to acquire high-dimensional data such as remote sensing image efficiently and accurately has become a research hotpot recent years. Deep learning is a kind of learning method which uses many kinds of simple layers to learn the mapping relation of complex layers. The authors will attempt to apply the deep belief network model (DBN), which is important in deep learning, to remote sensing image recognition. Using the new large-scale remote sensing image data set with abundant changes as the research object, the hierarchical training mechanism of DBNs is studied and compared with CNNS, the results show that the accuracy and speed of DBNs is better than that of CNNS, and more effective information can be obtained.