An Improved Hyperspectral Classification Algorithm Based on Back-Propagation Neural Networks

In this paper, a new method is proposed to improve the classification performance of hyperspectral images by combining the principal component analysis (PCA), genetic algorithm (GA), and artificial neural networks (ANNs). First, some characteristics of the hyperspectral remotely sensed data, such as high correlation, high redundancy, etc., are investigated. Based on the above analysis, we propose to use the principal component analysis to capture the main information existing in the hyperspectral images and reduce its dimensionality consequently. Next, we use neural networks to classify the reduced hyperspectral data. Since the back-propagation neural network we used is easy to suffer from the local minimum problem, we adopt a genetic algorithm to optimize the BP network´s weights and the threshold. Experimental results show that the classification accuracy is improved and the time of calculation is reduced as well.