Research On Multiple Classifier Combination Method Based On OLI Images

Remote sensing image classification is a key link in the process of remote sensing information extraction. It can provide geographic information and knowledge for researchers, as well as government regulators use classification information decision-making reference is put forward. High precision of remote sensing image classification technique has been the pursuit target in remote sensing application, how to improve the classification accuracy has become a key problem. Integrated thought and the method of integration of multiple classifiers has long been a research, in recent years, the use of multiple classifier combination for target recognition in the field of pattern recognition has been widely used, but in the field of remote sensing is in its infancy.OLI image is excellent, band set reasonable, widely used at present. This study based on the analysis of remote sensing image classification, especially the study of multiple classifier integration technology based on combing review, select OLI image as a classification study on the experimental data, the multiple classifiers in ENVI/IDL platform based on performance recognition matrix of voting, voting, maximum probability category method, comments the consistency principle combination method, and based on the sample clustering classifier selection method for the extraction of remote sensing information, the study area to study the combination of multiple classifiers related basic theory knowledge, through concrete experiment of multiple classifier compared with single classifier analysis, verified the multiple classifier combination of remote sensing classification accuracy improving the effectiveness of the method for multiple classifier after the practical application of laid a good foundation.The results show that:(1) To Kappa statistics, entropy, and diversity and precision index difference measure(D_P) as a classifier, can be the optimal classifier combination form;(2) VRPM, VMLC, DCS three multiple classifier combination algorithm can get high classification accuracy than single classifier.(3) The classification of multiple classifier combination method is helpful to improve the efficiency of classification of remote sensing application.