Agricultural Land Extraction In Nanjing Based On Time Series GF-1 WFV Images

Urbanization is one of the most important indicators of social economic development,but at the same time,it is likely to compromise national food security and objectives of sustainable strategy.As the unique mega city in the Yangtze River Delta,Nanjing city has to step up strict demands for the spatial regulations of land use due to its high urbanization rate,thus promoting the development of modernized precision agriculture oriented by information characteristics.The information of farmland scale and spatial pattern can be effectively extracted using high-resolution remote sensing images.In addition,time series high-resolution remote sensing images are very helpful to achieve precise extraction of farmland cultivating information together with dynamic monitoring of agricultural environment.The development of modernized agricultural dynamic monitoring based on remote sensing techniques promotes the urban agriculture to informationize,precision and sustainability in Nanjing.However,to date,there are few studies on accurate farmland extractions using GF-1 image,especially on the effective techniques and methods.Therefore,we proposed a new method herein to extract fine information of agricultural land based on GF-1 WFV images.Main research includes following four parts:(1)Based on the asymmetric Gaussian function fitting method,we reconstructed the features of land use and land cover type in a way that preserved the features while eliminating the noises.The reconstructed time series features curve is obviously smoothed.The typical features of bare soil,water,built-up area,forest and paddy field are easy to recognize.The differences between each type are clear to see,which fits the demand for fine extraction of agricultural land information.The reconstructed time series features curve lays a good foundation for further classification and finer extraction.(2)A hierarchical method for agricultural land use extraction based on reconstructed time series features is proposed.Firstly,the classification features such as high-resolution image spectral features and normalized water body index are used as preliminary classification features.Secondly,object-oriented shape features,radar VV/VH polarization bands and other features are applied to fine extraction of water and aquatic land,greenhouses,buildings,wheat and rapeseed.The overall classification accuracy and Kappa coefficient were 95,3 1%and 0.943 1,respectively,which achieved acceptable results.(3)Based on the results of fine extraction of agricultural land,the total agricultural land area of all districts in Nanjing was calculated and the Multiple Cropping Index(MCI)of crops in each district was derived.Through a comparative analysis of the statistical yearbook data,it is found that the overall MCI in Nanjing is 154.9%,and those of all agricultural regions are above 110%.The MCI of Liuhe District,Lishui District and Gaochun District is higher than the other regions.This paper analyzed the reason why the MCI in different regions of Nanjing are different from one another in terms of climate,national policy and agricultural mechanization level.We found that that the MCI interrelate with policies and mechanization levels,in addition to being affected by climate.Overall,time-series GF-1 WFV images can greatly contribute to the fine extraction and monitoring of agricultural land.The proposed method also potentially benefits the remote sensing technique and promote the intensive agriculture.