The Study of AdaBoost Algorithm to Enhance Support Vector Machine Application on Urban Rice Land Classification

摘要

Support vector machine (SVM) algorithm is to transform low dimensional feature space into high dimensional feature space to find out the maximum division margin between classes. SVM is widely used as a classifier of remote sensing images. However, the quality of training data may affect the accuracy of the classified image result. Boosting algorithm is a kind of ensemble learning which has the property of examining the training samples and enhances the classification accuracy. Boosting is a statistics method and can give the wrong samples a heavier weight and reduce the right samples' weight, then via iterative estimation to decrease the error rate of training samples. In this study, we use the high dimension advantage of SVM to classify satellite image. And combine AdaBoost algorithm to enhance SVM classifier. Discussing the results of AdaBoost integrate with SVM in satellite image classification in different complexities study area. The final result shows that the accuracy of AdaBoostSVM was increased 6.04% in higher complexity study area. AdaBoost ensemble learning can improve the commission error of paddy rice which caused form similar vegetation spectrum reflection. Besides, AdaBoost has better classification capability in the ridge between paddy rice fields. The result in this study proved that AdaBoost algorithm integrates with SVM can enhance and improve the classification of high resolution satellite image in complicated urban area.