Search

Soybean breeders must develop early-maturing, standard, and late-maturing varieties for planting at different latitudes to ensure that soybean plants fully utilize solar radiation. Therefore, timely monitoring of soybean breeding line maturity is crucial for soybean harvesting management and yield measurement. Currently, the widely used deep learning models focus more on extracting deep image features, whereas shallow image feature information is ignored. In this study, we designed a new convolutional neural network (CNN) architecture, called DS-SoybeanNet, to improve the performance of unmanned aerial vehicle (UAV)-based soybean maturity information monitoring. DS-SoybeanNet can extract and utilize both shallow and deep image features. We used a high-definition digital camera on board a UAV to collect high-definition soybean canopy digital images. A total of 2662 soybean canopy digital images were obtained from two soybean breeding fields (fields F1 and F2). We compared the soybean maturity classification accuracies of (i) conventional machine learning methods (support vector machine (SVM) and random forest (RF)), (ii) current deep learning methods (InceptionResNetV2, MobileNetV2, and ResNet50), and (iii) our proposed DS-SoybeanNet method. Our results show the following: (1) The conventional machine learning methods (SVM and RF) had faster calculation times than the deep learning methods (InceptionResNetV2, MobileNetV2, and ResNet50) and our proposed DS-SoybeanNet method. For example, the computation speed of RF was 0.03 s per 1000 images. However, the conventional machine learning methods had lower overall accuracies (field F2: 63.37–65.38%) than the proposed DS-SoybeanNet (Field F2: 86.26%). (2) The performances of the current deep learning and conventional machine learning methods notably decreased when tested on a new dataset. For example, the overall accuracies of MobileNetV2 for fields F1 and F2 were 97.52% and 52.75%, respectively. (3) The proposed DS-SoybeanNet model can provide high-performance soybean maturity classification results. It showed a computation speed of 11.770 s per 1000 images and overall accuracies for fields F1 and F2 of 99.19% and 86.26%, respectively. [ABSTRACT FROM AUTHOR]

Timely and accurately estimates of crop biomass and grain yield estimation are crucial for agricultural management. Optical remote sensing techniques can provide crop parameters (e.g., biomass, fractional vegetation cover (FVC)) at regional and larger scales. However, such techniques saturate at high crop canopy cover and cannot detect biomass stored in reproductive organs. The AquaCrop model can be used to estimate FVC, biomass, and grain yield output based on crop growth environmental parameters (e.g., temperature, rainfall, irrigation). In this work, we developed a method for estimating and mapping crop biomass and grain yield using unmanned aerial vehicle (UAV) remote sensing images and AquaCrop. The combination of low-cost UAV remote sensing data and AquaCrop can be used to map wheat biomass and grain yield before harvest. This work investigated whether biomass and grain yield can be predicted using measurements of biomass and FVC in several winter-wheat fields at an early growing stage using ground-based and remote sensing-based methods. FVC, biomass, and grain yield were measured at several key growing stages (jointing (S1), heading (S2), flowering (S3), grain filling (S4)) and harvest (H) during 2014–2015. We specifically evaluated the performances of using field- and UAV-based FVC and biomass measurements from different growing-stage combinations (e.g., S1–S4, S1–S3, and S1–S2) to calibrate the AquaCrop model and estimate biomass and grain yield. The results indicate that winter-wheat biomass and grain yield can be estimated by calibrating AquaCrop using (i) biomass and (ii) biomass and FVC. The results also reveal that the biomass and grain yield estimated using FVC and AquaCrop have poor accuracy compared with the biomass and grain yield estimated using (i) only biomass and (ii) biomass and FVC. The results suggest that the combined use of UAV remote sensing and AquaCrop can be used to obtain maps of biomass and biomass yield. When estimating winter-wheat biomass and grain yield one month (13 May) before harvest (11 June), the predicted biomass and grain yield agreed with the measured values (biomass: n = 96, coefficient of determination (R2) = 0.61, mean absolute error (MAE) = 1.69 t ha–1, root-mean-square error (RMSE) = 2.10 t ha–1, normalized RMSE (nRMSE) = 18.5%; grain yield: n = 48, R2 = 0.63, MAE = 0.96 t ha–1, RMSE = 1.16 t ha–1, nRMSE = 21.9%). When estimating winter-wheat biomass and grain yield one and a half months (26 April) before harvest (11 June), the predicted biomass and grain yield was lower than when estimating winter-wheat biomass and grain yield one month before harvest (biomass: n = 144, R2 = 0.55, MAE = 2.52 t ha–1, RMSE = 3.34 t ha–1, nRMSE = 23.6%; grain yield: n = 48, R2 = 0.34, MAE = 1.68 t ha–1, RMSE = 2.17 t ha–1, nRMSE = 35.7%). [ABSTRACT FROM AUTHOR]

The crop above-ground biomass (AGB) is critically important for monitoring crop growth, and its accurate estimation can be used by agricultural managers to improve farmland management and to predict crop grain yield. Many studies have shown that models for the estimation of AGB for multiple crop growth stages based on optical remote sensing spectral indices (SIs) often underestimate the crop AGB in later growth stages due to saturation problems. The purpose of this study was to estimate winter-wheat AGB using (i) high-frequency information obtained from the image wavelet decomposition (IWD) of unmanned aerial vehicle (UAV)-based digital images of winter-wheat canopy (i.e. high-frequency IWD information), and (ii) variables obtained from the continuous wavelet transform (CWT) of hyperspectral images of winter-wheat canopy (i.e. hyperspectral CWT variables). Digital and hyperspectral images were acquired using a digital camera and an imaging spectrometer, both mounted on a UAV. Unlike optical SIs, high-frequency IWD information and hyperspectral CWT variables may not be limited by saturation problems for high winter-wheat canopy cover. Our results indicate that high-frequency IWD information and hyperspectral CWT variables both increase (decrease) with increasing winter-wheat AGB. A multiple linear stepwise regression technique was used to analyse the performance of (1) SIs, (2) CWT, (3) IWD, (4) SIs + CWT, (5) SIs + IWD, (6) CWT + IWD, and (7) SIs + CWT + IWD, for estimating AGB. Our results indicate that: (i) the method based only on optical SIs may not support the estimation of winter-wheat AGB for multiple growth stages (SIs: coefficient of determination (R2) = 0.62, mean absolute error (MAE) = 1.30 t ha–1, root-mean-square error (RMSE) = 1.63 t ha–1); (ii) the combined use of the IWD of digital images of winter-wheat canopy and the CWT of hyperspectral images of crop canopy can support the estimation of winter-wheat AGB for multiple growth stages (CWT + IWD: R2 = 0.85, MAE = 0.79 t ha–1, RMSE = 1.01 t ha–1), including later growth stages; and (iii) the combined use of SIs and the IWD of digital images of winter-wheat canopy can improve the estimation accuracy of winter-wheat AGB (SIs + IWD: R2 = 0.80, MAE = 0.93 t ha–1, RMSE = 1.22 t ha–1), which may indicate that imaging spectrometers and cheap digital cameras have distinct advantages and can both be used to obtain AGB estimates using different methods. This work provides a new perspective on the use of high-frequency IWD information, hyperspectral CWT variables, and their combination to estimate AGB for multiple crop growth stages. [ABSTRACT FROM AUTHOR]

Land-cover classification provides the crucial data component for related Earth-science research. Currently, although multiscale remote-sensing images are the main source of data for classifying land-cover, the response of multi-resolution images to land-cover classification remains highly uncertain. In addition, because of the scarcity of time-synchronous multiscale satellite images of certain regions, up-scaling algorithms are generally used to generate and apply multiscale images. However, when using resolution-resampling images, it remains uncertain to what extent spectral loss or information distortion is responsible for the underlying differences in the accuracy of land-cover classification of various landscapes. To clarify this situation, we study the Hetian basin of Changting County in Fujian Province, south-east China by using quasi-synchronous multiple-resolution satellite observations (seven spatial resolution levels: 1 m, 2 m, 4 m, 8 m, 16 m, 30 m, and 50 m) to investigate possible correlations between spatial resolution and the land-cover classification. The classification is obtained by applying a support vector machine spectral classifier to random recordings made in 1875 sample plots. We also explore the effect of using lower-resolution images by comparing the classification results obtained by using several common up-scaling algorithms, such as nearest neighbour (NN), bilinear (BI), cubic convolution (CC), and pixel aggregation (PA). The results indicate that classification accuracy is significantly influenced by the spatial resolution of images (p < 0.05), with the accuracy increasing as the spatial resolution goes from 1 m to 4 m, then decreasing as the spatial resolution decreases beyond 4 m. In addition, for a resolution of 1 m to 30 m, almost all the up-scaled images provide a classification accuracy that differs from that obtained by using the native remote-sensing images of each resolution (p < 0.05), and the difference increases as the spatial-resolution ratio or up-scaling amplification factor increases. According to an analysis of the spatial scale of images using, e.g., multiband spectral reflectance and vegetation index, the up-scaling algorithms are less sensitive to spatial resolution and represent poorly the actual image characteristics. This result is attributed to the strong dependence of the spectral information in up-scaled images on the original images, which leads to discrepancies with respect to actual observations at the given scale. These results indicate that the effects of resolution cannot be ignored and that resampling data may not be adequate for multi-spatial-scale classification compared with the native satellite images. It is thus urgent to obtain an effective up-scaling algorithm that sharply reduces the problems caused by spatial heterogeneity. [ABSTRACT FROM AUTHOR]