The Method And Application Of The Information Extraction Of Vegetation Based On ZY3Images

Forest as a kind of natural geographical landscape,is constituted by the natural geographicfeatures including forest vegetation, and shows the comprehensive characteristics of theelements. Since the1970s, remote sensing is used to forestry in China, because of itsmacroscopic, comprehensive, dynamic, and low cost etc, has become the research methods ofideal for the study of forest resource status and dynamic changes. Forest spectral information ofremote sensing data is not only characteristic of forest vegetation spectral information, but thecomprehensive characteristics of forest landscape. Forest is the main body of forest vegetationand its change is the main embodiment of forest change. Therefore, the extraction of vegetationinformation is particularly important. In recent years, with the wide application of highresolution remote sensing image data, resource monitoring of forest vegetation have moreprecise information sources, it is convenient for the forest resources research of a small scale byremote sensing data.In this research we use remote sensing images of ZY3of Anji County of ZhejiangProvince as main date sources. It is mainly about vegetation information extraction ofpixel-based classification and object-oriented classification,with the help of remote sensingsoftware such as ERDAS9.2,eCognition8.0，ArcGIS9.3,etc.in order to find the method ofvegetation information extraction in mountainous hilly region and analyze the change of thebamboo forest in the study area according to the result of information extraction. The mainconclusions are as follows:(1)The Chinese ZY3satellite has2.1m resolution of panchromatic and5.8m resolution ofmulti-spectral. It has good effect of using in forest especially in the mountainous hilly regionsuch as the study area of this paper. And it has greater advantage on data acquisition ofdomestic satellite, especially domestic region coverage is higher, as well as the imageacquisition costs is much lower relatived to foreign satellites of the same quality, can replaceforeign satellite data very well.(2)In this paper，we used different methods of remote sensing image classification such aspixel-based classification and object-oriented classification,with texture information,vegetation, water, architecture characteristics of the remote sensing image as auxiliaryclassification information. Combined part of the data of the second class survey in Anji County,analyzed the overall accuracy and Kappa coefficient. The method of object-orientedclassification in mountainous hilly region on the vegetation information extraction has higherprecision and is more convenient and accurately on change extraction.(3) In object-oriented classification method, image segmentation is the key to theclassification. Different scales of segmentation have a large impact on the results of classification. In this paper, four different scales of segmentation including50,100,150,200have been experiment. The segmentation scale of100is found the best scale to extractclassification information of vegetation, which can effectively extract different vegetationclasses of study area. The overall classification accuracy is92%; Kappa coefficient is88.12%;the accuracy of each individual classification categories are more than80%, especially inextraction characteristics information of bamboo in study area. The user accuracy is96.45%andproduction accuracy is96.00%. When the segmentation scale is100, the number of objects inthe division generated is about20,000, and the split time is less than20minutes, which requiresthe modest hardware and has extensive feasibility in practice.(4) According to the calculation of magnitude and of various types of land use in the studyarea, the change area of bamboo shows a positive growth during the study period of2007to2012. The changing magnitude and dynamic degree is4.38%and0.88%. The decrease of otherkinds of forest contributes to the area growth of bamboo forest. The area of bamboo forest is723.13km2in2012, and approximately164.42km2come from other kinds of forest land,accounted for about22.74%of the total area of bamboo. The reduction of non-forest land alsomakes an important contribution to the increase the total area of bamboo forest, approximately25.01km2have transformed to bamboo.