Research On Forest Vegetation Classification By Remote Sensing In The Northern Of Liuhe County Jilin Province

Remote sensing automatic classification is the foundation of application, which is concerned by many scholar, and its has more and more application areas. At present, remote sensing survey of forest resources is generally rely on artificial visual interpretation, and the accuracy of the results are relatively low, can not meet the requirement. Therefore, how to improve classification accuracy by new technologies of remote sensing automatic classification to meet user,s requirement is of great significance.In the course of the study, using a hierarchical classification of thinking, considering the appropriate time of image for classification of vegetation and the size of spatial resolution, I extract the cover of forest firstly, and then classify the forest vegetation within the cover. In the process of forest vegetation extracting, two methods are used. One is extracted by NDVI of ETM+ directly, another method is supervised classification which is by the fusion image of NDVI and RVI relay on ETM+(may,2002) and band three of SPOT5(August,2007) , the latter method has a more higher precision, and is used as the basement of classification. With the three network model provided by IDRISI software, the forest vegetation is classified into six study area categories, then the three results are compared and analyzed. Test shows that these three artificial neural network classification have a very high prospect, on the whole have a complex mapping capabilities, especially the classification accuracy based on FYZZY-ARTMAP is relatively high, its Adaptability, and convergence is also an advantage, and its kappa coefficient reaches 0.8179.This paper is composed of five parts. In the first chapter, firstly it introduces the background of the research, and then summarizes the research progress of new technologies in remote sensing automatic classification at home and abroad, finally introduces the content and significance of this research. In the second chapter it analyzes and evaluates the traditional and the new methods that commonly used in remote sensing classification of forest vegetation, as well as the advantages, disadvantages and applicable conditions of these new technologies. The third chapter presents an overview of the study area, as well as GIS and remote sensing data preparation process and methods, and then get the best band of classification through analyzing the characteristics of remote sensing image. In the Part IV of this paper introduces the categories of vegetation classification, and then three artificial neural network models were used in the classification of forest vegetation in the study area, finally makes an analysis and evaluation on research results. Part V is the conclusion and prospects.