Search

Aboveground biomass (AGB) is a critical indicator for monitoring the crop growth status and predicting yields. UAV remote sensing technology offers an efficient and non-destructive method for collecting crop information in small-scale agricultural fields. High-resolution hyperspectral images provide abundant spectral-textural information, but whether they can enhance the accuracy of crop biomass estimations remains subject to further investigation. This study evaluates the predictability of winter canola AGB by integrating the narrowband spectra and texture features from UAV hyperspectral images. Specifically, narrowband spectra and vegetation indices were extracted from the hyperspectral images. The Gray Level Co-occurrence Matrix (GLCM) method was employed to compute texture indices. Correlation analysis and autocorrelation analysis were utilized to determine the final spectral feature scheme, texture feature scheme, and spectral-texture feature scheme. Subsequently, machine learning algorithms were applied to develop estimation models for winter canola biomass. The results indicate: (1) For spectra features, narrow-bands at 450~510 nm, 680~738 nm, 910~940 nm wavelength, as well as vegetation indices containing red-edge narrow-bands, showed outstanding performance with correlation coefficients ranging from 0.49 to 0.65; For texture features, narrow-band texture parameters CON, DIS, ENT, ASM, and vegetation index texture parameter COR demonstrated significant performance, with correlation coefficients between 0.65 and 0.72; (2) The Adaboost model using the spectra-texture feature scheme exhibited the best performance in estimating winter canola biomass (R2 = 0.91; RMSE = 1710.79 kg/ha; NRMSE = 19.88%); (3) The combined use of narrowband spectra and texture feature significantly improved the estimation accuracy of winter canola biomass. Compared to the spectra feature scheme, the model’s R2 increased by 11.2%, RMSE decreased by 29%, and NRMSE reduced by 17%. These findings provide a reference for studies on UAV hyperspectral remote sensing monitoring of crop growth status.

The task of converting facial sketch images to facial images aims to generate reasonable and clear facial images from a given facial sketch image. However, the facial images generated by existing methods are often blurry and suffer from edge overflow issues. In this study, we proposed HE-CycleGAN, a novel facial-image generation network with a symmetric architecture. The proposed HE-CycleGAN has two identical generators, two identical patch discriminators, and two identical edge discriminators. Therefore, HE-CycleGAN forms a symmetrical architecture. We added a newly designed high-frequency feature extractor (HFFE) to the generator of HE-CycleGAN. The HFFE can extract high-frequency detail features from the feature maps’ output, using the three convolutional modules at the front end of the generator, and feed them to the end of the generator to enrich the details of the generated face. To address the issue of facial edge overflow, we have designed a multi-scale wavelet edge discriminator (MSWED) to determine the rationality of facial edges and better constrain them. We trained and tested the proposed HE-CycleGAN on CUHK, XM2VTS, and AR datasets. The experimental results indicate that HE-CycleGAN can generate higher quality facial images than several state-of-the-art methods.

Drip fertigation (DF) is a widely used technology to increase grain yield with water and fertilizer conservation. However, the mechanism of high grain yield (GY) under DF is still unclear. Here, a four-year field experiment assessed the impacts of four treatments (i.e., conventional irrigation and nitrogen application, CK; drip irrigation with conventional nitrogen fertilization, DI; split-nitrogen fertigation with conventional irrigation, SF; and drip fertigation, DF) on maize phenology, leaf photosynthetic rates, grain filling processes, plant biomass, and GY. The results showed that DF significantly increased maize GY by affecting phenology, grain filling traits, aboveground biomass (BIO) accumulation, and translocation. Specifically, DF significantly increased leaf chlorophyll content, which enhanced leaf photosynthetic rates, and together with an increase of leaf area index, promoted BIO accumulation. As a result, the BIO at the silking stage of DF increased by 29.5%, transported biomass increased by 109.2% (1.2 t ha−1), and the accumulation of BIO after silking increased by 23.1% (1.7 t ha−1) compared with CK. Meanwhile, DF prolonged grain filling days, significantly increased the grain weight of 100 kernels, and promoted GY increase. Compared with CK, the four-year averaged GY and BIO increased by 34.3% and 26.8% under DF; a 29.7%, 46.1%, and 24.2% GY increase and a 30.7%, 39.5%, and 29.9% BIO increase were contributed by irrigation, nitrogen, and coupling effects of irrigation and nitrogen, respectively. These results reveal the high yield mechanism of drip-fertigated maize, and are of important significance for promoting the application of drip fertigation.

The leaf area index (LAI) is a crucial physiological indicator of crop growth. This paper introduces a new spectral index to overcome angle effects in estimating the LAI of crops. This study quantitatively analyzes the relationship between LAI and multi-angle hyperspectral reflectance from the canopy of winter oilseed rape (Brassica napus L.) at various growth stages, nitrogen application levels and coverage methods. The angular stability of 16 traditional vegetation indices (VIs) for monitoring the LAI was tested under nine view zenith angles (VZAs). These multi-angle VIs were input into machine learning models including support vector machine (SVM), eXtreme gradient boosting (XGBoost), and Random Forest (RF) to determine the optimal monitoring strategy. The results indicated that the back-scattering direction outperformed the vertical and forward-scattering direction in terms of monitoring the LAI. In the solar principal plane (SPP), EVI-1 and REP showed angle stability and high accuracy in monitoring the LAI. Nevertheless, this relationship was influenced by experimental conditions and growth stages. Compared with traditional VIs, the observation perspective insensitivity vegetation index (OPIVI) had the highest correlation with the LAI (r = 0.77–0.85). The linear regression model based on single-angle OPIVI was most accurate at −15° (R2 = 0.71). The LAI monitoring achieved using a multi-angle OPIVI-RF model had the higher accuracy, with an R2 of 0.77 and with a root mean square error (RMSE) of 0.38 cm2·cm−2. This study provides valuable insights for selecting VIs that overcome the angle effect in future drone and satellite applications.

Remote sensing (RS) scene classification has always attracted much attention as an elemental and hot topic in the RS community. In recent years, many methods using convolutional neural networks (CNNs) and other advanced machine-learning techniques have been proposed. Their performance is excellent; however, they are disabled when there are noisy labels (i.e., RS scenes with incorrect labels), which is inevitable and common in practice. To address this problem, some specific RS classification models have been developed. Although feasible, their behavior is still limited by the complex contents of RS scenes, excessive noise filtering schemes, and intricate noise-tolerant learning strategies. To further enhance the RS classification results under the noisy scenario and overcome the above limitations, in this paper we propose a multiscale information exploration network (MIEN) and a progressive learning algorithm (PLA). MIEN involves two identical sub-networks whose goals are completing the classification and recognizing possible noisy RS scenes. In addition, we develop a transformer-assistive multiscale fusion module (TAMSFM) to enhance MIEN’s behavior in exploring the local, global, and multiscale contents within RS scenes. PLA encompasses a dual-view negative-learning (DNL) stage, an adaptively positive-learning (APL) stage, and an exhaustive soft-label-learning (ESL) stage. Their aim is to learn the relationships between RS scenes and irrelevant semantics, model the links between clean RS scenes and their labels, and generate reliable pseudo-labels. This way, MIEN can be thoroughly trained under the noisy scenario. We simulate noisy scenarios and conduct extensive experiments using three public RS scene data sets. The positive experimental results demonstrate that our MIEN and PLA can fully understand RS scenes and resist the negative influence of noisy samples.