| With the continuous development of surveying and mapping satellite technology in the world and China,surveying and mapping satellite has the functions of sub-meter spatial resolution,high radiation accuracy and high positioning accuracy.Model function of surveying and mapping satellite with different difference,surveying and mapping data are also different,so the surface data will also be differences in the instruments by the atmosphere and the mirror of the radiation spectrum of remote sensing image reflected by the surface object data,there are many differences of land use change research difference results in the study area,aksu city policy execution the farmland on returning farmland to forest and grass in recent years,residential areas and has been expanding its traffic,increasing population,urbanization process inevitable,aksu city in the use of land resources in northern great changes have taken place in breadth and longitudinally,reasonable control of land resources,Accurate understanding of the process of urban expansion requires remote sensing data to clearly and prominently reflect the utilization status of land resources.Therefore,the timeliness and effectiveness of remote sensing vector data are of great significance to the relationship between land use change and urban expansion change in aksu city.Aiming at the technology development and application expansion of remote sensing image to reduce redundant information to extract land use information,the spatial scale problem,which mainly studies scale conversion,scale effect and optimal scale selection of various ground features,has become an important research direction of remote sensing image classification.In this paper,gaofen I image and Landsat8 OLI image from 2014 to 2018 were used as data sources,and the north of aksu city was taken as the research area to conduct methodological research on the scale problem of classification of gaofen I image and Landsat8 OLI image.Firstly,scale conversion of remote sensing image data after the expansion(fusion)of small-scale high-resolution images is carried out to obtain 15 m spatial resolution images and scale conversion effect evaluation,so as to obtain the method with the best scale conversion effect.Secondly,the series images after scale conversion are calculated to analyze the appropriate expression scale of the images in the research area.Five typical land feature samples,including cultivated land,construction land,forest land,unused land and water,were selected to calculate the separability between samples,and the optimal expression scale of typical land feature was obtained.Finally,using this method,Landsat8 OLI remote sensing image data were used to conduct multi-scale remote sensing classification in the north of aksu city.The main research contents and conclusions are as follows:(1)scale conversion method.Scale conversion of gaofen I and Landsat8 OLI remote sensing images in the research area was carried out by cubic convolution method,and the texture features of scale conversion images and original image texture features were evaluated and analyzed by using four indexes: contrast,correlation,average gradient Angle second-order distance and information entropy.The results show that the texture features of remote sensing images obtained by cubic convolution scaling are superior to the original images obtained by other methods,no matter Landsat8 OLI images or high-resolution no.1 images.(2)selection of optimal scale.By calculating the average local variance of classes in different regions,the spatial resolution of the scale conversion data from 2m to 30 m is used to select the optimal scale for classes in different regions.The results showed that the optimal resolution between cultivated land and unused area was 2m.When the spatial resolution of construction land,forest land and water is 15 m,the separability is better.(3)multi-scale classification.Neural network classification method was used to classify the scale conversion data of GF-1 images,the 2m spatial resolution images of GF-1 original images,Landsat8 OLI 15 m and 30 m spatial resolution images,and to evaluate the overall accuracy and kappa coefficient of all classification results.Conclusion: the accuracy of image classification results with a resolution of 15 m obtained through scale conversion is significantly higher than that of Landsat8 OLI images with a spatial resolution of 15 m,and the resolution is GF-1 image classification results.According to the accuracy evaluation and analysis of the classification results of series images after scale conversion,the overall effect of the classification resultsof transformed image data is good,and the resolution of 16 m can be the best resolution for the extraction of typical ground features in the research area.(4)land use change prediction in the north of aksu city.Based on GF-1 image transform the data of the scale of the image,using the markov prediction model,in2019 and 2020,to simulate the change of land use are: construction land,cultivated land,vegetation,2019 will present growth trend,unused showed a trend of decrease,water stable,basic in 2019 to 34.12% of the total area of north of aksu city construction land,arable land accounts for 30.24% of the total area,unused accounted for 29.28%,water accounted for 2.13% of total area,accounting for 4.23% of the total area of vegetation.According to the result data,the construction land area in 2019 is2031.16km2,and the cultivated land area is 1806.53km2,and the unused land area is1732.46km2,and the vegetation surface is connected with 125.45km2,and the water area is 250.91km2.In 2020,the construction land area will be 2020km2,the cultivated land area will be 2077.75km2,the unused land area will be 1885.39km2,the vegetation area will be 259.61km2,the construction land will account for 34.78%of the total area in the north of aksu city,the cultivated land will account for 31.56%of the total area,the unused land will account for 27.12%,the water area will account for 2.19% of the total area,and the vegetation will account for 4.35% of the total area. |
PDF Full Text Request