Research On The Phenological Phase Identification Of Winter Wheat In Multipolarization Spaceborne SAR Images

Agricultural production is the basis of the survival and development of human society and the basis and guarantee of all social activities.As an extremely important food crop,the production capacity of winter wheat is of great significance for ensuring food security,maintaining social stability and economic development.With the rapid development of remote sensing technology,remote sensing technology can quickly and efficiently obtain large-scale observation data to the ground,and can realize crop monitoring,disaster warning,yield prediction and reasonable planning of planting area.Spaceborne synthetic aperture radar(SAR)is a kind of microwave imaging sensor,which is not affected by cloud,rain and fog.It has all day and all-weather imaging ability.It can realize large-scale and long-time crop monitoring,and timely and accurately grasp the details of each growth stage of crops.Therefore,it is of great significance to identify and monitor winter wheat phenology based on Spaceborne SAR images.So far,many teams at home and abroad have used SAR images to identify Crop Phenology,especially rice phenology,as one of the most important food crops in the world,wheat is still in its infancy and needs further research.In order to solve the above problems,the identification method of winter wheat phenology based on multi polarization SAR image is proposed by using time series Aentinel-1A and Radarsat-2 images,and the phenological identification of Winter Wheat Based on time series bipolar and total polarization SAR images is realized respectively.The main contents of this paper can be summarized as follows:Scattering characteristics of winter wheat phenology based on time series Sentinel-1A and Radarsat-2images.In this study,the backscattering coefficients,polarization decomposition parameters and texture features of time series SAR images are analyzed by fully considering different polarization modes,different polarization combinations and different time phase SAR data features.Combining the results of polarization parameter analysis with the characteristics of winter wheat phenology,the scattering characteristics of winter wheat phenology are analyzed.The results show that the SAR image characteristics will show a certain periodic law with the change of winter wheat phenology.The phenological period of winter wheat can be divided into five stages:early nutrition stage,hibernation stage,growth stage,reproductive stage and mature stage.In the early stage of nutrition,the backscattering coefficient is larger due to single bare soil surface scattering;in hibernation period,the backscattering coefficient remains relatively stable;in growth period,reproductive period and mature period,with the increase of winter wheat plant,the backscattering coefficient increases,resulting in an overall trend of enhancement.The polarization decomposition parameters scattering entropy(H)and scattering angle(α)will increase and reach the peak value with the increase of winter wheat plants from the early vegetative stage to the growth stage,and will decrease at the mature stage due to the decrease of water content of wheat plants and other factors.Based on time series dual polarization Sentinel-1A images,a method of winter wheat phenology recognition based on decision tree was proposed.Firstly,the time series data are preprocessed to extract the backscattering coefficient of winter wheat;then the eigenvalue H/A/αdecomposition method is used to decompose the data by polarization,and the scattering characteristics of Winter Wheat in different phenological stages are analyzed by scattering entropy(H),scattering angle(α)and anisotropic(A)polarization decomposition parameters,and the decision tree is determined by combining with the box diagram analysis results In order to realize the recognition of winter wheat phenology,the optimal threshold was used.The experimental results show that the recognition accuracy is 67%in the reproductive stage and87%in the mature stage.The probability of misclassification increases because the growth state of winter wheat is similar between the growth stage and the reproductive stage.The overall accuracy of this method is79%,which has important practical significance for the study of winter wheat phenology recognition.Based on time series full polarimetric Radarsat-2 images,a decision tree method based on different polarimetric decomposition parameters was proposed to recognize different phenological phases of winter wheat.Firstly,the time series data are preprocessed to extract the backscattering coefficient;then the entropy(H),scattering angle(α)and anisotropy(A)are extracted by the H/A/αdecomposition method;the surface scattering(P_S),volume scattering(P_V),secondary scattering(P_D)and spiral scattering(P_H)are extracted by the Yamaguchi decomposition method;finally,the polarization parameters of H,a andαand the polarization parameters ofαand PS are extracted respectively Two decision tree methods were proposed to identify the winter wheat phenology.The experimental results show that the recognition accuracy of the decision tree method based on H/A/αpolarization decomposition parameters is 82%,and the recognition accuracy of the decision tree method based on H/A/αpolarization decomposition and Yamaguchi decomposition parameters is 86%.Surface scattering and volume scattering can improve the recognition accuracy of growth phase,and reduce the misclassification probability of growth phase and reproductive phase.Compared with the recognition accuracy based on H/A/αdecomposition parameter,the recognition accuracy is improved by 4%,and compared with the recognition accuracy based on dual polarization Sentinel-1A image,the recognition accuracy is improved by 7%.