The Study On Method Of Extracting Paddy Rice Area Based On SAR Image Data

Hainan Province belongs to the southern region of China.The remote sensing monitoring of rice using optical remote sensing data is often disturbed by clouds,fog and rain.It is impossible to obtain effective long time series optical remote sensing data and to identify and extract paddy rice accurately.Although many studies focus on cloud removal of optical remote sensing data,fusion of multi-source remote sensing data and smoothing of phenological curve,which can reduce the impact of cloud noise,it can not eliminate the impact of cloud noise fundamentally,thus limiting the accuracy of paddy rice identification and extraction.In this study,synthetic aperture radar(SAR)data has the advantages of all-weather and all-time.However,there are a series of problems such as purchase data and increasing cost for GF-3 and other multi-polarization synthetic aperture radar data.Therefore,this study chooses the Sentinel-1A radar data with long time series and high resolution,which can be obtained free of charge,to carry out a detailed study around the scientific issue of identifying and extracting the planting area of early paddy rice.In order to improve the accuracy of identification and extraction of early paddy rice planting area,and to make the same polarization data or the same polarization data at the same time better reflect the characteristics of backscattering coefficient in time domain.Considering that the backscattering coefficient of water body in time domain is relatively stable and low in each period,the normalized ratio of the pre-processed polarimetric SAR data is processed by using the idea of polarimetric differential SAR image and polarimetric ratio SAR image.Using expert knowledge decision tree classification method(threshold classification method based on sample statistical analysis)to realize the classification and recognition of early paddy rice region and non-early paddy rice region,the feasibility and application potential of Sentinel-1A radar data to extract early rice planting area in southern region were discussed,and the water normalization parameters were discussed to improve the quality of polarimetric SAR image and increase the planting area of early paddy rice.The advantage of extraction accuracy.The results and conclusions of this study are as follows:(1)The extracted early paddy rice planting area is basically consistent with the main early paddy rice production base in Hainan Province.Early paddy rice planting areas mainly distribute in the northwest,northeast and north-central areas of Hainan Province,accounting for more than 70% of the whole province’s early paddy rice planting area,and the rest mainly distribute in the coastal areas around Hainan Island.From confusion matrix analysis,area analysis,visual interpretation analysis and accuracy evaluation analysis at the municipal and county levels,it is concluded that Hainan Province has a good effect in identifying and extracting early rice planting area.Overall Accuracy(OA)was 91.25%,Kappa coefficient was 0.8167,mapping accuracy was 83.60%,user accuracy was 78.97%.The area of early rice reached 144.70 hectares,which was 14.55 thousand hectares,relative error was 6.38%,compared with the statistical data of 130.15 thousand hectares of early rice in Hainan Province in 2017.(2)The best monitoring coefficient for extracting early paddy rice planting area is multi-temporal NDVH data.NDVH data can effectively reduce the influence of polarization mode,coherent speckle noise and topographic factors on SAR image,and can effectively improve the quality of SAR image and highlight the texture and tone characteristics of early paddy rice region.The results show that the multi-temporal NDVH data is more suitable for identifying and extracting the planting area of early paddy rice in Hainan Province,and can more effectively simulate the real growth and development phenological characteristics of early paddy rice.(3)The best classification method is the threshold classification method based on expert knowledge decision tree.Supervised classification method and expert knowledge decision tree classification method are used to extract polarimetric SAR data after pretreatment.The results show that supervised classification method and CART algorithm are not suitable for polarimetric SAR data extraction in terrain fluctuations and complex areas.Only threshold classification method based on sample statistical analysis is more suitable for polarimetric SAR data extraction in Hainan.Study on the planting area of early paddy rice in Henan Province.(4)The study was Combining Sentinel-2 optical data and DEM data,the accuracy of early paddy rice recognition and extraction can be improved.Because of the complex terrain and large topographic fluctuation in the study area,the polarimetric SAR echo signal has a large deviation.According to this study,only using polarimetric SAR data and Normalized parameters of water body can not reduce this kind of impact.Therefore,combining the advantages of optical data and DEM data in this respect,it can effectively assist polarimetric SAR data to reduce the impact caused by topographic factors.(5)The long time series and high resolution SAR image data have certain advantages over medium and low resolution optical images(Landsat,MODIS).It can effectively simulate the phenological characteristics of early paddy rice growth and development in cloudy and rainy Hainan Province,make up for the shortcomings of optical images,and further reduce the influence of mixed pixels.The innovation of this research: In this study,a new identification and extraction parameter for monitoring early paddy rice planting area based on polarimetric SAR data was proposed by processing the water normalized ratio of the pre-processed polarimetric SAR data using the free available Sentinel-1A radar data.