Research of remote sensing classification about land survey based on SVM

Traditional classification algorithms used in remote sensing images have many problems, such as the low operation speed, low accuracy and difficult convergence. Support Vector Machine (SVM) is a new machine learning method of statistical learning theory based on small samples of machine learning rules. This paper deals with the remote sensing image classification by the support vector machine, using land cover information for classification of SPOT high spatial resolution images. Analysis was conduct ed on comparison of this method with tradition method. The result shows that the remote sensing image classification based on SV M method can solve the image classification fragmentation, low accuracy etc, and has advantage in study speed, orientation ability and expression, etc. This method has good prospects of application in both classifier training time and classification time.