粒子群优化算法在全极化SAR影像非监督分类中的应用

摘要

The high dimensionality of Polarimetric SAR image and the large volume which is resulted by the high resolution make the complexity of PolSAR image classification increasing. As a new evolutionary computation algo-rithm Particle swarm optimization PSO has powerful ability of making a global optimization. A classification method of PolSAR image based on PSO algorithm is studied in this paper. Firstly an initial classification based on scattering mechanism is done by taking use of the H/αmethod in the method. And then this classification result is used to initialize PSO algorithm. Finally the PSO is applied to output classification result of PolSAR data iterative-ly. The L-band PolSAR image of NASA-JPL laboratory and X-band high resolution PolSAR image of China Elec-tronic Technology Group were applied in the classification experiments. And the result shows that the classification precision of the PSO-based algorithm is higher than the H/α-Wishart algorithm.%极化SAR影像高维数以及高分辨率带来的大数据量特点使得影像分类的复杂度不断增加。粒子群优化( PSO)算法作为新型进化计算技术，具有强大的全局寻优能力。本文研究了一种基于PSO算法的极化SAR影像的分类方法。该方法首先利用H/α方法对数据进行基于散射机理的初分类；然后利用分类结果对PSO算法进行初始化；最后采用PSO对极化SAR数据迭代分类。实验采用NASA-JPL实验室的极化SAR数据以及中国电子科技集团 X 波段原型样机的高分辨率数据。结果表明，H/α-PSO 分类方法较H/α-Wishart分类精度及目视效果均有所提高。