Utilization Of Multispectral Images Segmentation Technology In Desertification Prevention And Cure

The classification for land cover of TM multispectral Imaging is a difficult problem all the time. Because there are all kinds of land covers and different kinds of land covers will change along with the time's process and space's distribution continuously, land covers'reflection, multispectral image data will alter too. In addition, mutispectral data itself have the feature of"the same substance, the different composition; the same composition, the different substance"Data will also be noised during collection process. All disadvantages we talk above make some traditional method of classification perform poorly in aspect of computational complexity, classification accuracy, generalization performance.But the desert terrain mulitspectral images has an obvious feature of texture .So, the method which combines the SVM with the texture feature is proposed after analysing the main method of classification and the digital feture of the multispectral images can solve the problem perfectly.By analysing the results of the experiments, we found out the effect of SVM is much better than others. The traditional methods such as Bayesian method and Neural Networks are all based on the idea of experience minimize risk. SVM is based on the theory of structural risk minimization.It is mainly used in the context of lacking samples,and can solve many problems in a large extent such as model choosing problems, nonlinear problems, local minimum points problems.By taking account of the texture feature of the images, we combined it with SVM. The result of the experiments showed that this method has a higher accuracy.