POLSAR Image Classification Using BP Neural Network Based On Quantum Clonal Evolutionary Algorithm

POLSAR image classification is an important part of POLSAR image processing. And it is also a key technology of POLSAR image interpretation. POLSAR image classification is a typical example which extracts forefeet of interpreting system as specific application. A fast and accurate POLSAR image classification is a premise of realizing some actual applications such as target detection and recognition.In this paper, a BP neural network classifier based on quantum clonal evolutionary algorithm (QCEA) is proposed and it can be used for POLSAR image classification. There are two steps in this classifier: initial weights and thresholds of BP neural network are optimized by QCEA; weights and thresholds are trained by gradient descent algorithm.In this paper, the development of POLSAR image classification at home and abroad method are introduced firstly. Secondly, feature extractions of POLSAR image are explained and feature selection by parallel multi-criteria feature selection algorithm based on link-like agent genetic algorithm(LAGA) of POLSAR image are discussed in detail. Thirdly, origin, basic theory and procedure of QCEA are discussed, and how to optimized initial weights and thresholds of BP neural network with QCEA is proposed. Finally, features selected are input into BP neural network optimized by QCEA and the classication result is attained.Therefore, there are two important tasks in this paper. The first one is QCEA is used for optimizing initial weights and thresholds of BP neural networks. And this type of BP neural network can converge to global optimum value. The second one is BP neural networks optimized by QCEA is used for POLSAR image classification. The experimental results showed that the classification result of this algorithm is better than other algorithm and the classification accuracy is higher than other algorithm.