Study On Remote Sensing Of Land Cover Classification Based On Object-oriented Method

Because of the human activity and climate factors are developing in recentyears, land cover has changed dramic enormously. For quick acquiring and capturingaccurately land cover information, and for utilizes the land resource in a better way, Agreat number of studies are analyzed by Many scholars, at home or abroad. With theremote sensing technology development, high spatial resolution remote sensing data areused commonly in this field. Depending on its abundant spectrum and texture andspatial information, high spatial resolution remote sensing data are widely applied inmany different fields, for territorial, military, environmental monitoring, agriculture,surveying, mapping and traffic. Due to the increase in resolution, texture and spatialinformation are increased of the ground-objects, different spectra in the same object arealso increased. When use the traditional classification method to process it, it may causea serious problems in the classification such as noise.To solve the above-mentioned problems, based on the platform of Ecognition andused the data of ZY-102C/Landsat8/DEM, this paper analyzed the classification ofremote sensing image with high resolution by the method of Object Oriented inRinbung County, Xigaze Prefecture, Tibet. The results are as follows:First, preprocessing of the remote sensing data in the study area, mainly containGeometrical Correction and Image registration of the ZY-102C data, and mergetogether multispectral bands of ZY-102C data and panchromatic band of landsat8 image. The results indicated that the fusion image have large standard errors andabundant information, this is a good result to be used for this study.Second，Make an appropriate Classification System in the study area by acquiredetailed knowledge of the land cover situation in the field, and also by practical fieldinvestigation in part of this field. Different segmentation scale and different parameterused to this study may have different results, so we choose the best segmentation scaleand parameter after many experimentsThird, make full use of the spectrum, texture and geometry form of the differentfeatures, and then make an appropriate classification rules to complete the informationextraction of the remote sensing data in this study field.Fourth, this paper also used the traditional method based on pixel to get theclassification result, and compare to the result of Object-Oriented Methods, it’s showthat the Object-Oriented Methods used to acquire the land cover is better than thetraditional method, and the accuracy is higher.Last,Used patch analyst software for the study area land cover analysis oflandscape, Selected12landscape analysis Index, The results show that: all kinds oflandscape area allocated a larger difference, uniformity is not high; grasslandlandscape is dominant landscape of the region, can be seen as a matrix in the region,alpine meadows and grasslands and barren lands crushing degree; the overall shape ofthe landscape is relatively simple, but there are certain irregularities and othercharacteristics.