加权变异粒子群BP神经网络在遥感影像分类中的应用

Application of the Weighted Variable Particle Swarm BP Neural Network in the Remote Sensing Image Classification

在经典的BP神经网络框架支撑下,利用加权变异粒子群算法使神经网络的训练更加科学,同时也更好地发挥了粒子群算法的优点,使其分类效果更加精准。实验后的分类结果表明,与改进之前的BP神经网络相比,总体精度和Kappa系数分别提高了0.108 3和0.138 3;与支持向量机、最大似然及最小距离等分类方法进行了对比,分类效果均优于以上方法。加权变异粒子群BP神经网络不仅可以实现遥感影像的高精度分类,对解决"同谱异物"和"异物同谱"现象也具有一定的作用。

In recent years, hybrid neural network begins to emerge gradually as a new neural network and plays a very good effect in remote sensing image classification. And the weighted variant particle swarm BP neural network is the one of them. This algorithm makes the neural network more scientific with the weighted variable particle swarm algorithm under the framework of classic BP neural network, and educes its better side at the same time to reach the better effect in classification. The results show that this method is better than BP neural network after the experiment. Its overall accuracy of the classification result is 0.108 3 higher than BP neural network and the Kappa coefficient of the classification result is 0.138 3 higher than BP neural network. And the method is compared with the support vector machine, the maximum likehood and the minimum distance classification method, its classification result is better than them. In addition, this study shows that the weighted variant particle swarm BP neural network can not only realize the remote sensing image classification with high accuracy, but also have a good effect on same object with different spectra and different objects with same spectrum.