Extraction Of Land Use And Change Information With Object-oriented Technology And High-resolution Remote Sensing Images

The extraction and application of land use/cover changes are very important for grasping and analyzing the development of the regional economy,the growth of population,urbanization,etc.Remote sensing has been an effective means for extracting land use/cover information with its strength of acquiring data.In this thesis,the main urban area of Tongxiang city was selected as the study area.At first,using object-oriented techniques and very high resolution remote sensing data,we analyzed the segmentation effects and change laws at different segmentation scales of 0.2,0.5 and 1 m high resolution data,and determined their optimal segmentation scales with ESP tool under e Cognition.Then we extracted land use information of 2011 and 2018 respectively with SVM classifier and analyzed the changes between 2011 and 2018.Finally,we subdivided residence commercial services,health care,industry,education,government institutions from construction land based on POI data and remote sensing image of 2018.Moreover,some suggestions for improvement were given according to the spatial distribution of education land.The main conclusions were as follow:(1)With the improvement of spatial resolution,the optimal segmentation scale becomes larger and the number of image objects increases after segmentation.At the same time,the changes in the optimal segmentation scale and the number of objects are affected by the image quality,spectral,and texture information and other factors.Under the condition of equal band weight,the shape factor and compactness are 0.5 and 0.4 for 0.2 meter resolution image,0.3 and 0.5 for 0.5m resolution image,0.4 and 0.5 for 1m resolution image.And an optimal segmentation scale of the 0.2 meter remote sensing image of three level is 82,116 and 148,49,83 and 113 for 0.5 meter remote sensing image,45,65 and 104 for 1 meter remote sensing image.(2)The classification results of 2011 and 2018 were obtained respectively by using the SVM classifier based on training and validating samples,and the classification results were satisfactory with accuracy evaluation of validation samples.Through the analysis of land use dynamic degree of change and transfer matrix,the land use of the study area changed significantly from 2011 to 2018.The construction land was the dominant change type,the area of road and water area also increased during this period,while the area of cultivated land and garden land decreased.With the acceleration of urbanization and industrialization,the expansion of new urban areas and the construction of the Tongxiang Economic Development Zone and Wutong Industrial Park were the main cause for increasing of construction land.(3)Within the scope of construction land of 2018,in conjunction with the POI data and remote sensing image,extraction of residential land,commercial service land,education land,industrial land,health care land,and government agency land was carried out with resulting areas of 993.58 hectares,105.23 hectares,156.93 hectares,1313.63 hectares,25.16 hectares,and 19.65 hectares respectively.Based on the corresponding public service requirements and spatial distribution of residential land,the spatial distribution of educational land was analyzed and suggestions for improvement were given.