Remote Sensing Images Of The Pulse Coupled Neural Network-based Automatic Target Recognition

As the third generation neural network, Pulse Coupled Neural Network(PCNN), which especially suits image processing, imitates biology neural network better compared with traditional neural network. To apply PCNN to image processing, is currently a main focus in the research area, but there are some difficulties, therefore it is necessary to proceed farther research in the area.On the other hand, the automatic recognition technology of remote sensing military targets has great application meaning in high-tech war. It is also difficult to research because of the imaging quality and the complexity of targets' characters.The PCNN theory and its application to image processing are researched deeply. New Double-Level Parallelized Firing Pulse Coupled Neural Network (DLPFPCNN) model is brought forward, and its working principium and application areas are stated in detail. The new model is designed to solve segment difficulties in segmenting images with low contrast, low SNR and uniform gently changing background. DLPFPCNN is used to segment images of different fields, achieving better results compared to normal method and usual modified methods, which proves the availability of DLPFPCNN.Then DLPFPCNN is applied to segment water area in remote sensing images with bridge targets over water and port targets, preparing for the recognition next. Satisfied results are gained in experiments, proving that DLPFPCNN suits remote sensing image process. And then over water bridges targets and port targets are recognized combining their line character and knowledge. In addition, a new method to judge the optimal results automatically is brought forward, aiming at the characteristics of water area in remote sensing images, and satisfied results are gained.