Extracting of soil salinization by SVM and accuracy evaluation based on texture characteristic.

It is of great importance to obtain the earth surface property timely and effectively,which can help us to know the relationship between human and nature phenomena and to make decision.In this paper,the author takes the North Oasis of the Tarim Basin(the delta Oasis of Weigan and Kuqa Rivers) for example,in virtue of ENVI software and using ETM+ data;discussing the method of extracting of soil salinization information.Remote sensing image classification is an important and complex problem.But conventional remote sensing image classification methods are mostly based on Bayes' subjective probability theory;it has the shortcomings of the low classification accuracy,classification efficiency and high indeterminacy,because there are many defects,a new tendency is that the SVM has been applied to remote sensing image classification.This paper reports the classification method based on support vector machines(SVM),and introduces the fundamental theory of SVM algorithm,then incorporating of spectrum and texture information.The results indicate that on the windows of 3×3,5×5,7×7,9×9,11×11,13×13,the precision of classification is very high,up to above 93 %,and when validating the result of classification,we have found that the result of classification by the SVM method based on spectrum and texture information on 9×9 windows is in accordance with the fact.So we can say that joining with the texture character can easily extract the soil salinization information(severely salinized soil,moderately salinized soil,slightly salinized soil),and increases its precision.Therefore,the classification method by SVM(Support Vector Machines) based on texture characteristic can be adapted to RS image classification and monitoring of soil salinization,furthermore provide a effective way for remote sensing information extraction.