Study On Large Area Land Cover Dynamic Monitoring Based On Remote Sensing Image

Land cover can reflect the human social economic activities, influnce the watercirculation and material balance. Land cover changes affect biochemical circulation, also cancause the natural environment change, it is closely related to human. So the research on thedynamic monitoring of land cover has been a hotspot in the field of and the moderngeography and global change study. As one of the means of land cover research, remotesensing technology can observe large area, and the data it gets has many characteristics, suchas timeliness, objectivity, comprehensiveness and comparability. It has become one of themost reliable and effective means in the land cover observation.There are some problems in studing the huge area land cover by using the RS image,such as huge image data, complex type of the land cover, and complex spectral features,which means it can hardly conform to the requirements of the classification result by using theusual-single way. In this paper, aiming at monitoring a regional spatial distribution of landcover by using RS image, we make the following research:1. The area and data analysisAnalyze the natural conditions of the study area, base on that we determine the regionland cover types of classification system. Choose two Landsat TM multi-spectral remotesensing image, combined with classification system and the spectral features, analyze thespectral band combination, interpreting mark as well as the texture feature.2. The land cover information extractionIn view of the large area of remote sensing image sample selection, this paper putsforward a sample reuse method. That is to choose the sample of each type in the common areaof the adjacent image, and save the samples, then reuse the samples in other image that sharethe same pixal location. On this basis, we established the sample reuse technology of process;In land cover information extraction process, we use layered extraction method, by using thelayer mask, we extract the water, artificial cover, cultivated land, vegetation extraction one byone. Use SVM classification method to do the classifiction. 3. The change of the type analysis and forecastingBased on the classification results, combined with regional characteristics in the studyarea, we analyze the land cover change of the study area, use the Markov model to predict thechange of land cover in the following10years.The main results of this study are: The land cover classification method we choose canreflect research area types, the comparison between results of two phase image classificationcan reflect the change of land cover types; The sample reuse method reduces the difficulty ofsample selection, reduce the workload of the sample collection, and improve the adjacentimage edge consistency in a certain degree; By using a single class of layered extraction maskclassification can avoid irrelevant features in the classification process influence on singletype of object, effectively improve the accuracy of the land cover classification; The result ofthe land cover change in the study area analysis and prediction of the reasonable landutilization in the region has a certain practical value.