Rice Extraction Research And Rice Crop Parameter Estimation Based On Full-polarization SAR Data

Rice is one of the major food crop in the world. How to obtain large-scale ricegrowth information,that is an important scientific problem. Our country needefficient scheme to monitor and guild the rice production. These years, optical remotesensing data are widely used in rice monitoring. However, rice is grown mostly intropical and subtropical areas with warm tropical climates, so there are plenty of rainand cloud in rice growth season. SAR(Synthetic Aperture Radar) instruments have allweather capability, which is reliable data for rice monitoring. Our unique dataconsisted of Radarsat-2microwave backscattering coefficients at HH,VH,HV,VVpolarization for two rice crop stages. Biological parameters of whole rice plant weremeasured at the same time in Dongqiao of Jiangsu. The main results are as follows:1. Analysis of backscatter signature of rice crop. Through statistical analysis, weconclude that: jointing stage of rice is the best stage for rice extraction and HH/VVcan enhance rice information and be extraction factor.2. Research of rice extraction. There are two methods which are used to classifyon Radarsat-2image: SVM and Neural network. SVM is better than Neural network,and HH/VV can improve classification accuracy.3. The inversion of rice biological parameter. The study analyzed microwavebackscatter signatures in time domain, then built a simple backscatter process model(the water cloud model) and regression models. HH water cloud model and HH/VVexponent model were optimal to simulate rice fresh biomass.4. Water cloud model and HH/VV exponent model were used to map ricebiomass distribution in rice area of SAR image.