FUSION ALGORITHM OF PIXEL-BASED AND OBJECT-BASED CLASSIFIER FOR REMOTE SENSING IMAGE CLASSIFICATION

This paper proposes a new method to fusion pixel-based classifier and object-based classifier to land cover classification. We choose the Boosting classifier as pixel-based classifier and choose the SVM classifier as object-based classifier. At first, one scene image is classified using Boosting classifier to acquire the labels of each pixel point in the image. Secondly, the same scene image is segmented, and then we cast a vote to each segmentation block, and select the label of the highest votes as the label of the segmentation block. Thirdly, the results of vote and the classification results of SVM classifier are fusion. By we apply the method to Landsat TM, ZiYuan3 and IKONOS images for land cover classification, compare the results of new approach with the results of only using the Boosting algorithms and only using the SVM algorithms. Experimental results show that the significant improvement in classification accuracy.