| In recent years,as China's economy has entered the new normal,China's urbanization process has accelerated.The rapid growth of urban population and the disorderly expansion of construction land have brought adverse effects on the natural environment,which makes urban development face great challenges.Urban land use information can more comprehensively and intuitively display the urban internal spatial structure,which is of great significance to solve the problem of urban development.With the rapid development of remote sensing technology,more and more high-resolution remote sensing images appear,which creates favorable conditions for rapid and accurate acquisition of ground information.The object-oriented classification method takes into account more classification features when extracting ground feature information,and can be combined with Geoscience Knowledge and other thematic features to make the classification process closer to human cognitive process.It has become one of the mainstream directions of land use information extraction research.However,when extracting the information of urban construction land,it is difficult to obtain the semantic function information of different land types only by relying on remote sensing images,which has certain limitations.With the development of information technology and knowledge engineering,the application of spatial big data is gradually increasing,which provides a new idea for the extraction of urban construction land information.Based on this,this study explores the selection method of the optimal parameters of multi-scale segmentation,constructs and optimizes the image classification feature space based on GF-2 and GF-6 images.Land use information is extracted based on multi-level classification system,and the accuracy is evaluated.The special information of construction land in the main urban area of Xindu District is extracted through spatial big data,and its spatial distribution characteristics are analyzed.On this basis,the spatial structure of land use in Xindu District is analyzed,and the urban functional zoning is realized.The main conclusions are as follows:(1)Comprehensively using the maximum area method and the ESP based optimal scale evaluation method,select the optimal segmentation scale of each land use type,select the homogeneity factors under different segmentation scales based on the control variable method,and establish the multi-level structure of image classification based on the segmentation scale,so as to lay the foundation for subsequent land use information extraction.The segmentation parameters of each level of GF-2 image are[179,0.5,0.6],[133,0.4,0.5],[65,0.5,0.5].[178,0.4,0.6];The segmentation parameters of each level of GF-6 image are[123,0.3,0.5];[61,0.6,0.5](2)Through the sampling of typical features,the differences of spectral characteristics,texture characteristics,index characteristics,geometric characteristics of different land use types are compared.According to the characteristics of land use classification,46 features are selected to construct the initial feature space,and the feature parameters are optimized by combining the difference analysis of feature target eigenvalues and SPM model.From the results,spectral features account for a large proportion of the classification parameters at all levels of the image.When spectral features cannot effectively distinguish features,texture features,exponential features and geometric features are further introduced.(3)Based on CART decision tree algorithm and random forest algorithm,land use classification is carried out for GF-2 image and GF-6 image respectively.The results show that the classification accuracy of GF-2 image is better than that of GF-6 image,and the classification accuracy of CART decision tree and random forest algorithm is also different.In general,the classification accuracy based on random forest algorithm is higher than that of CART decision tree algorithm.After correcting the extraction results of land use information by visual interpretation method,the overall classification accuracy is 95.16%and the total Kappa coefficient is 0.9320.(4)The construction land in the land use classification results is subdivided through POI data,and the commercial land,residential land,industrial and mining storage land,government organization land,transportation service station land,special land,medical and health land,education and scientific research land and sports land are extracted respectively.Assisted by Chengdu smart city space-time big data and cloud platform and"Tianditu"national geographic information public service platform,sample points are selected on Gaode electronic map to evaluate the accuracy of classification results.The results show that the extraction accuracy of special information of construction land based on spatial big data is high,the overall classification accuracy is 91.82%,and the total Kappa coefficient is 0.8980.(5)According to statistical analysis,compared with construction land,non-construction land accounts for a larger proportion of land use in Xindu District,as much as 64.24%of the total land area in the main urban area of Xindu District.The area order of each land use type in non-construction land is:wood land>cultivated land>traffic land>bare land>water area>facility agricultural land.The construction land covers an area of 38.47 km~2,accounting for 35.76%of the total land area in the main urban area of Xindu District.The area order of various construction land is:industrial and mining storage land>residential land>commercial land>education and scientific research land>transportation service station land>special land>medical and health land>sports land>government organization land.(6)From the perspective of the spatial structure of land use in the main urban area of Xindu District,there are a large number of plots of residential land,industrial and mining storage land,and their distribution is relatively concentrated.The distribution of land for public management,public service and commercial use is relatively scattered;The number of plots for special land and transportation service stations is small.Through urban functional zoning,the study area is divided into six categories:functional area dominated by residential land,functional area dominated by commercial land,functional area dominated by industrial and mining storage land,cultivated land protection area and ecological protection area. |
PDF Full Text Request