A new kind of Combination Classifier and its application in Classification of Remote Sensing Image

Aiming at the problem that each sub classifier has its own shortcomings in classification of remote sensing image, promote a new combination classifier based on the improved voting rules. The combination of different sub-classifiers is combined by cascading and parallel connection, and the voting rules based on prior knowledge are used to achieve the accurate classification of remote sensing images. Taking Yueyang TM remote sensing image as experimental research object, multiple classifiers combination method is applied to classify processing, and the processing result is compared with the result of single classifier. Through error matrix comparison, we can see that the accuracy of Kappa coefficients of multiple classifiers is better than that of single classifiers. Compared with classification results, multiple classifier's effect is better than single classifier in detail results. The research results show that the combined classifier is much better than the single classifier in the classification of remote sensing images, and has better expansibility, which has wider application prospects in the field of geographical condition monitoring.