Remote sensing image classification based on multiple classifiers fusion

There are many methods for Remote Sensing (RS) image classification at present. Different classifiers can obtain diverse accuracies for different RS images or different feature types. Now, the research on RS image classification is focusing on developing new classifiers. Little research has been conducted on making full use of the complementation of different classifiers which may obtain more precise result than single classifier. In the paper, a weighted multiple classifiers fusion method on abstract level was proposed. Firstly, five classifiers were selected to classify one TM image. Then the classification experiment based on weighted multiple classifiers fusion on abstract level and on measurement level was finished. At last, the comparison of classification precision of single classifiers and fusion classifiers was done. Test result proved that the classification accuracy based on fused classifier is higher than single classifier obviously.