Unmanned aerial vehicle (UAV)-based remote sensing (RS) possesses the significant advantage of being able to efficiently collect images for precision agricultural applications. Although numerous methods have been proposed to monitor crop nitrogen (N) status in recent decades, just how to utilize an appropriate modeling algorithm to estimate crop leaf N content (LNC) remains poorly understood, especially based on UAV multispectral imagery. A comparative assessment of different modeling algorithms (i.e., simple and non-parametric modeling algorithms alongside the physical model retrieval method) for winter wheat LNC estimation is presented in this study. Experiments were conducted over two consecutive years and involved different winter wheat varieties, N rates, and planting densities. A five-band multispectral camera (i.e., 490 nm, 550 nm, 671 nm, 700 nm, and 800 nm) was mounted on a UAV to acquire canopy images across five critical growth stages. The results of this study showed that the best-performing vegetation index (VI) was the modified renormalized difference VI (RDVI), which had a determination coefficient (R²) of 0.73 and a root mean square error (RMSE) of 0.38. This method was also characterized by a high processing speed (0.03 s) for model calibration and validation. Among the 13 non-parametric modeling algorithms evaluated here, the random forest (RF) approach performed best, characterized by R² and RMSE values of 0.79 and 0.33, respectively. This method also had the advantage of full optical spectrum utilization and enabled flexible, non-linear fitting with a fast processing speed (2.3 s). Compared to the other two methods assessed here, the use of a look up table (LUT)-based radiative transfer model (RTM) remained challenging with regard to LNC estimation because of low prediction accuracy (i.e., an R² value of 0.62 and an RMSE value of 0.46) and slow processing speed. The RF approach is a fast and accurate technique for N estimation based on UAV multispectral imagery.

Given its high nutritional value and capacity to grow in harsh environments, quinoa has significant potential to address a range of food security concerns. Monitoring the development of phenotypic traits during field trials can provide insights into the varieties best suited to specific environmental conditions and management strategies. Unmanned aerial vehicles (UAVs) provide a promising means for phenotyping and offer the potential for new insights into relative plant performance. During a field trial exploring 141 quinoa accessions, a UAV-based multispectral camera was deployed to retrieve leaf area index (LAI) and SPAD-based chlorophyll across 378 control and 378 saline-irrigated plots using a random forest regression approach based on both individual spectral bands and 25 different vegetation indices (VIs) derived from the multispectral imagery. Results show that most VIs had stronger correlation with the LAI and SPAD-based chlorophyll measurements than individual bands. VIs including the red-edge band had high importance in SPAD-based chlorophyll predictions, while VIs including the near infrared band (but not the red-edge band) improved LAI prediction models. When applied to individual treatments (i.e. control or saline), the models trained using all data (i.e. both control and saline data) achieved high mapping accuracies for LAI (R² = 0.977–0.980, RMSE = 0.119–0.167) and SPAD-based chlorophyll (R² = 0.983–0.986, RMSE = 2.535–2.861). Overall, the study demonstrated that UAV-based remote sensing is not only useful for retrieving important phenotypic traits of quinoa, but that machine learning models trained on all available measurements can provide robust predictions for abiotic stress experiments.

Unmanned aerial vehicle (UAV) and satellite data have considerable complementarity for platform inter-operability, data fusion studies, calibration and validation efforts, and various multiscale analyses. To optimize cross-platform synergies between field-deployable UAV and space-based satellite systems, an understanding of spectral characteristics and compatibility is required. Here, we present the assessment of spectral consistency, undertaking a pixel-to-pixel similarity assessment of co-registered reflectance maps using corresponding spectral bands from UAV and satellite multispectral imagery. A high-resolution centimeter-scale UAV-mounted MicaSense RedEdge-MX sensor is intercompared against variable-resolution multi-spectral sensors on-board PlanetScope, Sentinel-2 and Landsat 8 platforms. Sampling from within an urban environment that covers a range of both natural and man-made surfaces, we employ ground-based spectroradiometer data to evaluate pixel-level responses, using regression analysis and measurements of relative root mean square error (rRMSE) to assess for factors such as spatial and spectral misalignment. Using two radiometric correction approaches for the UAV data, we found that a vicarious radiometric correction was more accurate than a linear empirical line method, with the former improving rRMSE by between 1.6% and 20.11% when assessed against spectroradiometer measurements. Spectral band misalignment between the UAV and satellite sensors affected their spectral consistency, causing different reflectance values for the same object in the corresponding UAV and satellite bands, with the issue amplified over specific land-cover classes (e.g. grass in the red edge part of the spectrum). Using the standard deviation of a UAV-derived normalized difference vegetation index (NDVI) as a metric of spatial heterogeneity, larger differences between the UAV and satellite-based NDVI were observed for different ground features in response to both land-cover boundary and shadow effects. Interestingly, higher spatial heterogeneity did not necessarily lead to higher spectral inconsistencies. It was also determined that as spatial scale differences between the UAV and satellite platforms increased, the lower was the impact of geometric misregistration on their consistency. Indeed, the rRMSE between the reflectance values of the UAV-based spectral bands and the corresponding satellite imagery was smaller at lower resolution (e.g. Landsat 8) than higher resolution (e.g. PlanetScope). Overall, the study provides insight into the collective effect of spectral and spatial misalignments on the degree of spectral consistency that can be expected between UAV and satellite data, guiding robust radiometric intercalibration efforts and the potential for improved synergy and interoperability between UAV and satellite data.

Tools for robust identification of crop diseases are crucial for timely intervention by farmers to minimize yield losses. Visual diagnosis of crop diseases is time-consuming and laborious, and has become increasingly unsuitable for the needs of modern agricultural production. Recently, deep convolutional neural networks (CNNs) have been used for crop disease diagnosis due to their rapidly improving accuracy in labeling images. However, previous CNN studies have mostly used images of single leaves photographed under controlled conditions, which limits operational field use. In addition, the wide variety of available CNNs and training options raises important questions regarding optimal methods of implementation of CNNs for disease diagnosis. Here, we present an assessment of seven typical CNNs (VGG-16, Inception-v3, ResNet-50, DenseNet-121, EfficentNet-B6, ShuffleNet-v2 and MobileNetV3) based on different training strategies for the identification of wheat main leaf diseases (powdery mildew, leaf rust and stripe rust) using field images. We developed a Field-based Wheat Diseases Images (FWDI) dataset of field-acquired images to supplement the public PlantVillage dataset of individual leaves imaged under controlled conditions. We found that a transfer-learning method employing retuning of all parameters produced the highest accuracy for all CNNs. Based on this training strategy, Inception-v3 achieved the highest identification accuracy of 92.5% on the test dataset. While lightweight CNN models (e.g., ShuffleNet-v2 and MobileNetV3) had shorter processing times (<0.007 s per image) and smaller memory requirements for the model parameters (<20 MB), their accuracy was relatively low (~87%). In addition to the role of CNN architecture in controlling overall accuracy, environmental effects (e.g., residual water stains on healthy leaves) were found to cause misclassifications in the field images. Moreover, the small size of some target symptoms and the similarity of symptoms between some different diseases further reduced the accuracy. Overall, the study provides insight into the collective effects of model architecture, training strategies and input datasets on the performance of CNNs, providing guidance for robust CNN design for timely and accurate crop disease diagnosis in a real-world environment.

Commercially available digital cameras can be mounted on an unmanned aerial vehicle (UAV) for crop growth monitoring in open-air fields as a low-cost, highly effective observation system. However, few studies have investigated their potential for nitrogen (N) status monitoring, and the performance of camera-derived vegetation indices (VIs) under different conditions remains poorly understood. In this study, five commonly used VIs derived from normal color (RGB) images and two typical VIs derived from color near-infrared (CIR) images were used to estimate leaf N concentration (LNC). To explore the potential of digital cameras for monitoring LNC at all crop growth stages, two new VIs were proposed, namely, the true color vegetation index (TCVI) from RGB images and the false color vegetation index (FCVI) from CIR images. The relationships between LNC and the different VIs varied at different stages. The commonly used VIs performed well at some stages, but the newly proposed TCVI and FCVI had the best performance at all stages. The performances of the VIs with red (or near-infrared) and green bands as the numerator were limited by saturation at intermediate to high LNCs (LNC > 3.0%), but the TCVI and FCVI had the ability to mitigate the saturation. The results of model validations further supported the superiority of the TCVI and FCVI for LNC estimation. Compared to the other VIs derived using RGB cameras, the relative root mean square errors (RRMSEs) of the TCVI were improved by 8.6% on average. For the CIR images, the best-performing VI for LNC was the FCVI (R² = 0.756, RRMSE = 14.18%). The LNC–TCVI and LNC–FCVI were stable under different cultivars, N application rates, and planting densities. The results confirmed the applicability of UAV-based RGB and CIR cameras for crop N status monitoring under different conditions, which should assist the precision management of N fertilizers in agronomic practices.