Identification And Extraction Of Typical Crops Based On GF-1 Time Series Data

Shijiazhuang city(in Hebei province of China)is an important production base of grain and fruit in North China Plain.As the increase of the grain planting area and the irrigation extraction of groundwater,resulting in water resources shortage of Shijiazhuang and the surrounding area so that brought the serious influence to the ecological environment.Therefore,it is important to quickly obtain accurate and all-round crop planting area and the spatial distribution to estimate the amount of groundwater and protect groundwater resources,what ' s more,to estimate the amount of crop irrigation.In this study,the GF-1 WFV remote sensing image was used as the data source to do the spatial sampling of typical crops in the study area by Google earth and field sampling.So the NDVI time series data of different crops could be constructed.The Savitzky-Golay filtering method(S-G),Harmonic analysis of time series(Hants),Double Logic Curve method(D-L),Non-symmetric Gauss Function method(A-G)were used to reconstruct the NDVI time series data(smooth)and extract the growth curve of various crops,combined with the R language to obtain the key phenological period of different crops,the phenological characteristics and spectral characteristics of crops were combined to identify and extract the winter wheat,summer maize,rice and fruit of the study area in 2015.So the following conclusions can be reached:(1)The Harmonic analysis of time series(Hants)and Double-Logistic Curve fitting method are more suitable for the reconstruction of time series data of typical crops and the reconstructed curve has the higher consistency in the expression of vegetation phenology;the time series harmonic analysis method has the least disturbance to the original data;while the rest two methods have the worst effect on typical crop reconstruction.(2)The classification accuracy obtained by the combination of key phenological characteristics and spectral characteristics of crops is higher than that of single spectral features.The extraction area of typical crops in Shijiazhuang was extracted accurately.In the extraction area,the extraction accuracy of winter wheat was 85%,and the precision of summer maize extraction was about 88%.The precision of rice extraction was about 86%,and the accuracy of fruit extraction was about 83%.