Study On Classification Of Poyang Lake LU Based On Remote Sensing Technique

Land resource is the foundation of existence and development for human being. Real time monitoring land use status is the basis of rational management of land resource, and the important gist for planning agriculture or even the whole national economy. Possessing the advantages of macroscopic information, real time, periodicity and integrity, remote sensing make it possible that land use is monitored quickly, objectively and precisely.At present, there is a far-reaching use of Remote Sensing (RS) image in the field of land use research .How to improve the accuracy of interpretation in order to promote the utility of RS technology becomes a urgent problem in RS application. In this paper, using the Landsat TM data, supported by the software of Erdas Imagine 8.4, taking an example of the PoYang Lake land use, made material application with all kinds of classification methods, estimate the accuracy of all kinds of classification, explore and expect the classification method of land use.The paper includes four parts:Part 1: It is the part of introduction. The RS conception, current situation and trend by using RS inside and outside of country were simply introduced and the significance was set forth.Part2: The conception and principle of the RS classification were simply introduced. Artificial Neural Networks, unsupervised and supervised classification of the RS image methods are introduced by the numbers. In the supervised classification, the election and passivation of the AOI and confirmation of the prior probability are described, and the principle of the maximum likelihood classification based on Bayes rule in the paper is introduced.Part3: It gave a detailed introduction about the classification process of PoYang Lake Land use supported by Erdas Imagine 8.4 software with the resource of Landsat TM. The land use classification of the synthesizing image of TM7,5,1 and TM5,4,1, principal component analysis synthesizing image, Artificial Neural Networks weredone respectively, and evaluated the result of the classification with error matrix. The total classification accuracy were 82.31%, 85.07%, 90.00%, 91.73% respectively. Part 4: Based on part 3, the research result and imperfection were put forward.