The Study Of Methods And Comparison Of Accuracy In Remote Sensing Image Classification On Land Use Updating Survey

With the remote sensing being combined with geographic information system, much remote sensing data is more and more applied to various areas. Because of the large amount of information of remote sensing, how quickly, accurately and simply to get the interesting information from these massive data has become an important research direction. So how to identify a variety of surface features is a main objective through processing and interpreting on remote sensing image. Both the extraction of the spatial information, monitoring the dynamic changes of land, or creating thematic maps and building the database of remote sensing image and so, which can not do without classification. Remote sensing classification is the main method of the analysis of remote sensing data and very important research content in remote sensing image procession.Land is the basis for human survival and development. With the economy developing, more intensive land resources can be used. Therefore, to grasp the situation of land use is particularly important. The timeliness of remote sensing is very good, which can get the information on land use quickly and timely. Moreover, the resolution of remote sensing images is also rising with the development of remote sensing technology; so it can provide more details for land use situation surveying. Using computer to make the land use classification on remote sensing images, not only can improve efficiency, but also can support the spatial data mining and knowledge discovery in future.To strengthen land management, that land database was the latest state of real-time needed to be reassured in 2004. The second national land survey was launched in order to grasp the situation of land use. In this paper, taken Hunyuan County, Datong City, Shanxi Province as an example, the methods of the land use investigation classification on remote sensing image were studied from the theory and the practice. The main contents of this paper are as follows:(1) A variety of methods in the land use updating survey for remote sensing interpretation were expounded, including the visual interpretation classification, Classification based on statistical analysis and new methods for remote sensing image classification. And the methods of accuracy analysis were expounded in remote sensing image classification.(2) Combined with the study area, the methods of sample selection were discussed in detail in the study area.(3) Using these three categories including minimum distance classification, maximum likelihood classification and human-computer interaction classification methods, the image in the study area were classified for the land use. (4) Through the confusion matrix and Kappa coefficient, the three classification results were conducted an accuracy comparison and analysis using the standard data.The experiment showed that the highest accuracy methods was man-machine interaction classification, but its efficiency was too low; Supervised classification not only was efficient, but also could be ensure certain degree of accuracy. In practice, a reasonable classification can be chosen according to the actual situation.