Land Cover Classification In Northern Laos Using Image Zoning And Object Oriented Methods

Land use and land cover change(LUCC)is one of the core contents of the current global climate change research.As a basic data for land and resource management,it is an important content to understand the status and change of land cover.Our country now is almost blank in the high precision geographic information,because developed countries impose technical and information blockade on China.Although series of GF satellites are widely used in domestic resources and environment,it is a problem for Chinese researchers to think more about how to fully exploit the potential of the application of GF data in many directions.Therefore,this paper uses GF1 satellite images as the main data source to study the land cover classification in northern Laos,and makes the progress:1.Image zoning use relief amplitude.The topography of the northern Laos is complex,and the accuracy of land cover classification in hill and plain areas will be affected by it.This paper uses DEM to calculate relief amplitude to classify the GF1 image zoning into the hills and plains.The results show that this method is accurate in reflecting the detailed land cover information,and provide more accurate data for the following work.2.Object oriented method classification.The KNN and Bayes methods are used to classify the hills and plains respectively.The results show that both methods are appropriate with Kappa coefficient higher than 0.85,but Bayes method is better.By comparing with classification results using maximum likelihood(MLC)and ISODATA method,this paper shows that the object-oriented classification method is better,and the accuracy in hill is improved by about 7%,and the accuracy in plain is improved by about 4%.3.This paper analyzes the influence of samples number on classification accuracy.In this study,5 groups of training samples are set up--20,40,60,80,100.The results show that the number of samples has different influence on the classification accuracy.The classification accuracy tends to be stable in the hill areas after samples number is 60,and the accuracy in plain is higher after samples number reaches 40.4.This paper applies object oriented method to large scale land cover classification,then analyzes and evaluates its spatial pattern.