Comparative Studies, Remote Sensing Classification Based On Support Vector Machine

Support Vector Machine (SVM) is a new machine learning method which foundations have been developed by Vapnik (1995), under small finite samples, it based on statistical learning theory. It, found on the principle of structural risk minimum (SRM), manifests many priorities than other method used before in actual questions of finite training samples. In classification of remote sensing image, with the method of SVM classification, image data don't have to degrade the high dimension and the speed & the precision have greatly improved.The algorithm of sequential minimum optimization (SMO) divides quadratic programming optimization problem into a series of smallest possible problems which are easy to deal with. It treats with two data each time and these small quadratic programming problems are solved analytically, which avoids using a time-consuming numerical quadratic programming problem optimization as an inner loop. These features give the SVM more advantages.Traditional remote sensing classification method, such as supervised and non-supervised classification methods, are all based on spectrum feature space without applying with texture feature space. If we regard the texture feature space as classification reference, we can make use of multiple feature space to classify in effect.The paper applies with spectrum and texture feature space to classify finite remote sensing data by SVM method with SMO algorithm. Firstly realizes the SMO classifier to support three vector data and extracts the texture features from remote sensing map, secondly with manual method to select training samples and format the data for the classifier, finally compares the results separately classified by remote sensing and SVM classification methods and compares the results separately with spectrum feature space and spectrum + texture feature space classified by SVM classification method.