Name
Abstract | ||
|---|---|---|
This paper presents an inversion method based on neural networks (NN) to estimate rice biomass in a paddy rice field with fully polarimetric (HH, HV, VH, VV) measurements at S-band. The backscattering coefficients are measured by a ground-based scatterometer system during the rice growth period from May to September 2010. The rice growth parameters including biomass, leaf-area index (LAI) and canopy structure are collected by random sampling at the same time. Data analyses show that the multi-temporal backscattering coefficients are very sensitive to the changes of biomass, LAI, canopy height and stem density. We also find that multi-temporal observations are suitable for paddy detection in the early growth period, and co-polarimetric observations perform well for monitoring rice status in the late growth period. According to the field measurements, a rice growth model was established as the function of rice age. The model made the parameters more representative and universal than limited random measurements over a given rice field. The scatter model of rice fields was simulated based on Monte Carlo simulations. The input parameters in the scatter model were generated by the rice growth model. The simulation results of the scatter model were composed as the NN training dataset, which was used for training and accessing the NN inversion algorithm. Two NN models, a simple training model (STM) and a related training model (RTM), were applied to estimate rice biomass. The obtained results show that the root mean square error (RMSE = 0.816 Kg/m2) of the RTM is better than that of the STM (RMSE = 1.226 kg/m2). The results suggest that the inversion model is able to estimate rice biomass with radar backscattering coefficients at S-band. |
Year | DOI | Venue |
|---|---|---|
2014 | 10.1109/JSTARS.2013.2282641 | Selected Topics in Applied Earth Observations and Remote Sensing, IEEE Journal of |
Keywords | Field | DocType |
rice age,neural network,paddy detection,multitemporal observations,backscattering coefficients,random measurements,neural networks,rice biomass estimation,canopy structure,nn inversion algorithm,leaf-area index,monte carlo simulations,paddy rice field,s-band,learning (artificial intelligence),rice growth model,rice growth parameters,canopy height,ground-based radar scatterometer,radar backscattering data,data analysis,rice status monitoring,ad 2010 05 to 09,fully polarimetric measurements,stem density,rice biomass,scatter model,remote sensing by radar,data analyses,simple training model,early growth period,crops,field measurements,multitemporal backscattering coefficients,monte carlo methods,random sampling,rice growth period,inversion model,ground-based scatterometer system,copolarimetric observations,root mean square error,vegetation mapping,growth model,related training model,nn training dataset,neural nets,radar polarimetry,mean square error methods,radar backscattering coefficients,learning artificial intelligence | Radar,Monte Carlo method,Paddy field,Remote sensing,Backscatter,Mean squared error,Scatterometer,Sampling (statistics),Inverse transform sampling,Mathematics | Journal |
Volume | Issue | ISSN |
7 | 2 | 1939-1404 |
Citations | PageRank | References |
3 | 0.40 | 13 |
Authors | ||
4 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Mingquan Jia | 1 | 6 | 4.83 |
| Ling Tong | 2 | 6 | 2.50 |
| Yuanzhi Zhang | 3 | 269 | 39.70 |
| Yan Chen | 4 | 133 | 14.37 |