Dioscorea PolystachyaTurczaninow is one of the most famous traditional Chinese Materia Medica. However, there is lack of large-scale classification method which is crucial for its growth status monitoring and yield estimation. This study proposed a reliableDioscorea PolystachyaTurczaninow classification model based on spectral-temporal-spatial feature optimization using time-series Sentinel-2 data. Firstly, 16 mono-temporal classification models were developed using five vegetation indices (VIs) and random forest algorithm. Then, temporal feature optimization was conducted by identifying the most effective time phase combinations based on Sentinel-2 time-series normalized difference vegetation index (NDVI) data, gaussian mixture modeling algorithm and F1 score ofDioscorea PolystachyaTurczaninow in each mono-temporal model. Next, spectral features were optimized by replacing NDVI with the optimal VI corresponding to each time phase, thus constructing a multi-VIs-based time-series dataset. Finally, the spatial feature optimization was conducted using the three-dimensional convolutional neural network (3-D CNN) algorithm and the multi-VIs-based time series Sentinel-2 data. Following the comprehensive feature optimization, the finalDioscorea PolystachyaTurczaninow classification model was determined. The results found that Sentinel-2 data acquired during the rhizome enlargement stage played a crucial role in classifying theDioscorea PolystachyaTurczaninow. By using the optimized features, the classification model achieved theDioscorea PolystachyaTurczaninow F1 score of 95.00%, which improved by 11.49% compared to only using the time series NDVI data. This spectral, temporal and spatial feature optimization method also has the potential to the development of large-scale, dynamic, and accurate mapping for other crops.

The GF-1 wide field view (WFV) data have wide view angles ranging from 0° to 48°, which generate notable angular effects for earth surface monitoring. However, current angular correction method for GF-1 WFV data relies on bidirectional reflectance distribution function (BRDF) parameters derived from coarse resolution data, leading to limited correction accuracy. Therefore, this study aims to develop a general set of BRDF parameters for 16-m WFV data angular effect correction and improving the vegetation parameter estimation accuracy. Firstly, more than 40 high-quality GF-1 WFV data in the North China Plain and Northeast China regions covering the typical vegetation types were collected to construct BRDF parameters. This study took into account three vegetation types (cropland, grassland and forest) and the normalized differential vegetation index (NDVI) magnitude. Through the least square method, a set of BRDF parameters were estimated based on various NDVI levels. Then, the nadir reflectance was calculated to estimate leaf area index (LAI) and fractional vegetation cover (FVC). Finally, the evaluation of the corrected reflectance indicated that the developed BRDF parameters performed best for correcting angular effect of cropland, and followed by grassland. In addition, the validation indicated that the generated BRDF parameters effectively improved the LAI and FVC estimation accuracy.

Leaf area index (LAI) and fractional vegetation cover (FVC) are two essential vegetation parameters for ecological and climate studies. The Chinese Gaofen-1(GF-1) wide field view (WFV) satellite data is a valuable data source for LAI and FVC retrieval at high spatio-temporal resolution. Like its name, GF-1 WFV has very large view angle ranging from 0° to 48°, which can impact the accuracy of vegetation parameter retrieval. The primary aim of the study was to develop an angular effect correction (AFX-fix) method that can effectively normalize GF-1 WFV data. Our objective was to enhance the applicability of the corrected data in retrieving LAI and FVC. The AFX-fix method used angular index, anisotropy flat index (AFX), and a fixed set of bidirectional reflectance distribution function (BRDF) parameters. LAI and FVC were retrieved from the GF-1 WFV reflectance data using the PROSAIL model combined with a random forest method. Results showed that the accuracy of LAI and FVC retrieval in wheat and corn from angular corrected GF-1 WFV data was improved with a decrease in root mean square error (RMSE) by 0.66 for LAI and 0.03 for FVC compared to that based on the original data. We anticipate that this new method will help improve the performance of LAI retrieval of these crop types using WFV data.

Dioscorea polystachyaTurczaninow is an herbaceous vine plant distributed in China, and its rhizome named Chinese yam is a famous traditional Chinese medicine for treating diabetes and other diseases. However, a large region monitoring method of its growth status is lacking, which is important for Chinese yam yield estimation. Therefore, this study proposed a leaf area index (LAI) estimation algorithm forDioscorea polystachyaTurczaninow using a hybrid method and Sentinel-2 vegetation indices. First, two feature selection algorithms the gradient boosting regression tree (GBRT) and absolute Pearson correlation coefficient (APCC), were combined with field-measured data and radiation transfer model simulated data to generate four different feature important ranking groups. Then, a hybrid feature selection algorithm was used to determine the best feature subsets under each ranking group, and GBRT regression and least absolute shrinkage and selection operator (LASSO) were used to develop the LAI estimation models. Finally, the best LAI estimation model forDioscorea polystachyaTurczaninow was determined based on validation accuracy. The results indicated that the field-measured data were more reliable than the simulated data for feature selection, and the best LAI estimation model was the LASSO model using nine selected vegetation indices, which achieved the performance with RMSE of 0.391 and MAE of 0.310. The proposed method could provide real-time LAI estimates for future Chinese yam yield prediction.

Here, four normalized difference water index (NDWI) variants, i.e., NDWI(860,970), NDWI(860,1240), NDWI(860,1640), and NDWI(1240,1640)are generated from the corn-oriented PROSAIL radiative transfer model. It is found that, instead of the linear relationship derived in previous studies, corn canopy water content (CWC) is best approximated as an exponential function of NDWI. Following the analysis of the PROSAIL-generated results, a newly optimized NDWI-based scheme is proposed for estimating corn CWC according to variations in the performance of the four NDWI variants under different CWC conditions. Validation results based on independent field data from the SMEX02, HiWATER2012, and Baoding2018 field experiments verify that this optimized NDWI-based corn CWC estimating scheme has a higher accuracy ($R = 0.87\,\,\pm \,\,0.03$, RMSE = 0.2068 ± 0.0145 kg/m2) than existing NDWI-based strategies for corn CWC retrieval. The feasibility of retrieving corn vegetation water content (VWC) based on the optimized NDWI-based scheme is also investigated, and the superiority of the optimized NDWI-based scheme for retrieving corn VWC is assessed. By comparing with four other NDWI-based corn VWC estimating methods, as well as the corn VWC parameterization scheme applied in the SMAP soil moisture algorithm, it is shown that our optimized NDWI-based scheme has the best VWC estimation accuracy, with the highest$R$of 0.89 ± 0.02 and the lowest RMSE of 0.7179 ± 0.0555 kg/m2.

The matrix doubling (MD) model is a numerical solution to the radiative transfer equation. It can achieve better accuracy in simulating microwave signals from vegetated terrain by considering multiple-scattering effects. However, it is difficult to apply the MD model to retrieving work due to its high complexity. This letter presents a case study performed on corn to demonstrate a multiangular (5°-65°), multiband (1.4/6.925/10.65 GHz) microwave emission model considering multiple-scattering effects by parameterizing the MD model. The simulated emissivity differences between the theoretical model and parameterized model are small. The mean absolute percent errors are all less than 1%, and the root mean square errors (RMSEs) are all within the range of 10-3. Validations using airborne polarimetric L-band microwave radiometer data and ground-based trunk-mounted multifrequency microwave radiometer data indicate that the parameterized model achieves good accuracy with overall RMSEs within 8K at all three bands.

In this paper, we present a new approach to count the number of people that cross a counting line from video images. This paper focuses on point-level annotation in training images and incorporate spatial features along with novel temporal features in training the structured random forest for estimating crowd density. By computing the crowd velocity, we model the crowd counting map as elementwise multiplication of crowd density map and crowd velocity map. Integrating over crowd counting map on the line of interest(LOI) locations leads to the instantaneous LOI counting numbers. We show that results are comparable to those obtained when using more complex and costly techniques.