一种基于证据理论与神经网络的遥感影像分类方法

A Remote Sensing Image Classification Method Based on Evidence Theory and Neural Networks

把影像的空间信息融入分类决策 ,提出了一种基于证据理论与神经网络的遥感影像分类方法。对原图像作平滑处理 ,得到原图像的平滑图像 ;利用神经网络对原图像及其平滑图像分别进行训练、分类 ;利用证据理论对它们的分类结果 (决策 )进行融合 ;最后 ,把融合结果 (决策 )作为原图像的最终分类结果。实验结果与性能比较表明 ,新方法是有效的 ,提高了影像的分类精度。

Neural networks are widely used in remote sensing image classification. The spatial information of the image and evidence theory is applied to classification of remote sensing image based on neural networks. It can significantly increase classification accuracy. Firstly, the original image to be classified is smoothed to obtain a smoothed image. Secondly, artificial neural networks (ANNs) are used to train and classify the original image and its smoothed image, respectively. Thirdly, the two classification outputs of ANN are fused with evidence theory. Finally, the fused result is considered as the classification result of the original image. Experimental results show that the new method is very efficient, and the classification accuracy is greatly improved compared with the classic ANN method.