Dynamic Monitoring Of Urban Green Space Based On Granular Computing Remote Sensing Image Processing Technology

With the deepening of China’s urbanization process,a large number of green space has been planned as construction land,resulting in the continuous reduction of the self purifi-cation capacity of urban land,while the emissions of urban domestic waste and waste gas and sewage are increasing year by year,which seriously restricts the sustainable develop-ment of the environment.As an important means to ensure ecological livability,urban green space system has great value in ecological and social economic benefits.Remote sensing image processing technology provides favorable data support for urban green space dynamic change information acquisition.Aiming at the problem of poor classi-fication effect caused by large amount of remote sensing image data and high attribute dimension,this paper proposes an attribute reduction method based on optimistic multi-granulation rough set,combines with support vector machine classification model to clas-sify urban land types,and analyzes the classification results,and gives urban green space planning suggestions.Firstly,the data region is selected and preprocessed by image clipping,radiometric cal-ibration and image mosaic.Secondly,the visual interpretation method is used to label training samples,and the attribute decision table of training samples is obtained by at-tribute value calculation.Then,based on the proposed optimistic multi-granulation rough set attribute reduction algorithm,attribute reduction is carried out according to the char-acteristics of the training sample attribute decision table,and compared with the two at-tribute reduction-algorithms,the best reduction algorithm parameters are selected and the reduced attribute set is obtained.And after parameter tuning gets the best parameters,support vector machine is used to get classification of urban land types.Finally,accord-ing to the classification results and data analysis,the change of urban green space in the study area is obtained,and combined with the actual urban development,the urban green space planning suggestions are given.The numerical results of attribute reduction show that the proposed optimistic multi-granulation attribute reduction algorithm(RF-OMG)greatly improves the computational efficiency.In the five selected datasets,the average running time of RF-OMG is only 1.3%of CDG algorithm,and 0.6%of NFRS algorithm.On the training sample data set,RF-OMG can quickly identify key attributes and achieve the highest classification accuracy,and the running time is only 1.3%of NFRS.The classification results of the dynamic monitoring model show that the reduction rate of urban green space area of Changsha in the nine years from 2009 to 2018 is significantly slower than that in the first nine years of 2009,and the percentage of construction land converted to green space has increased from 0.56%in 2009 to 4.4%in 2018,which in-dicates that while the expansion speed of urban construction land is slowing down,more attention has been paid to the construction of urban green space.In terms of urban green space planning,it is suggested to build new artificial green space,pay attention to the development of urban riverside green space,and pay attention to suit measures to local conditions,and strengthen the later maintenance and management of urban green space.