Explore the vast range of titles available.

Small object detection in uncrewed aerial vehicle (UAV) imagery is hindered by limited pixels, insufficient detailed information, and strong background interference, leading to weak feature representation and poor contextual modeling. To address these issues, we propose a multi-domain enhancement and cross-layer feature fusion detection Transformer (MDCL-DETR) with progressive feature processing. First, a multi-domain enhancement module (MDEM) based on CSP (cross stage partial) structure is proposed, which fuses spatial and frequency-domain features in a lightweight manner to enhance object detail and global structures while effectively distinguishing object features from background interference. Second, a cross-layer feature extraction module (CLEM) is introduced to aggregate multi-scale features across layers, alleviate information loss caused by downsampling, and preserve spatial details of small objects while integrating high-level contextual semantics. Meanwhile, a gated Mamba fusion module (GMFM) is proposed, which adopts the Mamba architecture for long-range dependency modeling of multi-scale features and integrates a gating mechanism to realize the dynamic weighted fusion of local details and global context, further improving feature discriminability and global modeling capability. Finally, a fine-grained enhancement module (FGEM) is designed, which leverages feature reorganization and adaptive feature extraction to reinforce and compensate fine-grained features. Extensive experimental results validate the effectiveness and generalization of the proposed method, achieving mAP50 scores of 54.1% and 56.2% on the VisDrone2019 and AI-TOD datasets.

Maize (Zea mays L.) is one of the most consumed grains in the world. Within the context of continuous climate change and the reduced availability of arable land, it is urgent to breed new maize varieties and screen for the desired traits, e.g., high yield and strong stress tolerance. Traditional phenotyping methods relying on manual assessment are time-consuming and prone to human errors. Recently, the application of uncrewed aerial vehicles (UAVs) has gained increasing attention in plant phenotyping due to their efficiency in data collection. Moreover, hyperspectral sensors integrated with UAVs can offer data streams with high spectral and spatial resolutions, which are valuable for estimating plant traits. In this study, we collected UAV hyperspectral imagery over a maize breeding field biweekly across the growing season, resulting in 11 data collections in total. Multiple machine learning models were developed to estimate the grain yield and flowering time of the maize breeding lines using the hyperspectral imagery. The performance of the machine learning models and the efficacy of different hyperspectral features were evaluated. The results showed that the models with the multi-temporal imagery outperformed those with imagery from single data collections, and the ridge regression using the full band reflectance achieved the best estimation accuracies, with the correlation coefficients (r) between the estimates and ground truth of 0.54 for grain yield, 0.91 for days to silking, and 0.92 for days to anthesis. In addition, we assessed the estimation performance with data acquired at different growth stages to identify the good periods for the UAV survey. The best estimation results were achieved using the data collected around the tasseling stage (VT) for the grain yield estimation and around the reproductive stages (R1 or R4) for the flowering time estimation. Our results showed that the robust phenotyping framework proposed in this study has great potential to help breeders efficiently estimate key agronomic traits at early growth stages.

Recent research has highlighted the potential for high-resolution, high-density, 3D and 4D ground-penetrating radar (GPR) acquisitions on alpine glaciers. When carried out on foot, such surveys are laborious and time consuming, which limits their application to small domains of limited glaciological interest. Further, crevasses and other hazards make the data acquisition risky. To address these issues, we have developed a drone-based GPR system. The system has a payload weight of 2.2 kg and a data output rate of 14 traces per second. An 80-MHz antenna and a recording time of 2800 ns mean that depths of over 100 m can be reached in temperate ice. Differential GPS positioning assures accurate flight paths. At a speed of 4 m s−1 and height of 5 m above the glacier surface, our system can acquire over 4 line-km of GPR data in 20 min on a single set of drone batteries. After presenting the technical specifications of the system and tests required to optimize its performance, we showcase a recently acquired 3D dataset from the Otemma glacier in Switzerland, where 462 parallel GPR profiles were surveyed at a 1-m line spacing, totaling over 112 line-km of data, in only 4 days.

Precise positioning is critical for autonomous uncrewed aerial vehicle (UAV) navigation, especially in GNSS-denied environments where radio-based signals are unreliable. This study presents an optical camera communication (OCC)-based positioning method that enables real-time 3D coordinate estimation using aviation obstruction light-emitting diodes (LEDs) as optical transmitters and a UAV-mounted camera as the receiver. In the proposed system, absolute positional identifiers are encoded into color-shift-keying-modulated optical signals emitted by fixed LEDs and captured by the UAV camera. The UAV’s 3D position is estimated by integrating the decoded LED information with geometric constraints through the Perspective-n-Point algorithm, eliminating the need for satellite or RF-based localization infrastructure. A virtual offshore wind farm, developed in Unreal Engine, was used to experimentally evaluate the feasibility and accuracy of the method. Results demonstrate submeter localization precision over a 50,000 cm flight path, confirming the system’s capability for reliable, real-time positioning. These findings indicate that OCC-based positioning provides a cost-effective and robust alternative for UAV navigation in complex or communication-restricted environments. The offshore wind farm inspection scenario further highlights the method’s potential for industrial operation and maintenance tasks and underscores the promise of integrating optical wireless communication into autonomous UAV systems.

While Uncrewed Aerial Vehicle (UAV) systems and camera sensors are routinely deployed in conjunction with Structure from Motion (SfM) techniques to derive 3D models of fluvial systems, in the presence of vegetation these techniques are subject to large errors. This is because of the high structural complexity of vegetation and inability of processing techniques to identify bare earth points in vegetated areas. Furthermore, for eco-geomorphic applications where characterization of the vegetation is an important aim when collecting fluvial survey data, the issues are compounded, and an alternative survey method is required. Laser Scanning techniques have been shown to be a suitable technique for discretizing both bare earth and vegetation, owing to the high spatial density of collected data and the ability of some systems to deliver dual (e.g., first and last) returns. Herein we detail the development and testing of a UAV mounted LiDAR and Multispectral camera system and processing workflow, with application to a specific river field location and reference to eco-hydraulic research generally. We show that the system and data processing workflow has the ability to detect bare earth, vegetation structure and NDVI type outputs which are superior to SfM outputs alone, and which are shown to be more accurate and repeatable, with a level of detection of under 0.1 m. These characteristics of the developed sensor package and workflows offer great potential for future eco-geomorphic research.

Uncrewed aerial vehicles (UAVs) have become essential for remote sensing in extreme environments like Antarctica, but detecting moss and lichen using conventional red, green, blue (RGB) and multispectral sensors remains challenging. This study investigates the potential of hyperspectral imaging (HSI) for mapping cryptogamic vegetation and presents a workflow combining UAVs, ground observations, and machine learning (ML) classifiers. Data collected during a 2023 summer expedition to Antarctic Specially Protected Area 135, East Antarctica, were used to evaluate 12 configurations derived from five ML models, including gradient boosting (XGBoost, CatBoost) and convolutional neural networks (CNNs) (G2C-Conv2D, G2C-Conv3D, and UNet), tested with full and light input feature sets. The results show that common indices like normalised difference vegetation index (NDVI) are inadequate for moss and lichen detection, while novel spectral indices are more effective. Full models achieved high performance, with CatBoost and UNet reaching 98.3% and 99.7% weighted average accuracy, respectively. Light models using eight key wavelengths (i.e., 404, 480, 560, 655, 678, 740, 888, and 920 nm) performed well, with CatBoost at 95.5% and UNet at 99.8%, demonstrating suitability for preliminary monitoring of moss health and lichen. These findings underscore the importance of key spectral bands for large-scale HSI monitoring using UAVs and satellites in Antarctica, especially in geographic regions with limited spectral range.

Fractional vegetation cover (FVC) is an essential metric for valuating ecosystem health and soil erosion. Traditional ground-measuring methods are inadequate for large-scale FVC monitoring, while remote sensing-based estimation approaches face issues such as spatial scale discrepancies between ground truth data and image pixels, as well as limited sample representativeness. This study proposes a method for FVC estimation integrating uncrewed aerial vehicle (UAV) and satellite imagery using machine learning (ML) models. First, we assess the vegetation extraction performance of three classification methods (OBIA-RF, threshold, and K-means) under UAV imagery. The optimal method is then selected for binary classification and aggregated to generate high-accuracy FVC reference data matching the spatial resolutions of different satellite images. Subsequently, we construct FVC estimation models using four ML algorithms (KNN, MLP, RF, and XGBoost) and utilize the SHapley Additive exPlanation (SHAP) method to assess the impact of spectral features and vegetation indices (VIs) on model predictions. Finally, the best model is used to map FVC in the study region. Our results indicate that the OBIA-RF method effectively extract vegetation information from UAV images, achieving an average precision and recall of 0.906 and 0.929, respectively. This method effectively generates high-accuracy FVC reference data. With the improvement in the spatial resolution of satellite images, the variability of FVC data decreases and spatial continuity increases. The RF model outperforms others in FVC estimation at 10 m and 20 m resolutions, with R2 values of 0.827 and 0.929, respectively. Conversely, the XGBoost model achieves the highest accuracy at a 30 m resolution, with an R2 of 0.847. This study also found that FVC was significantly related to a number of satellite image VIs (including red edge and near-infrared bands), and this correlation was enhanced in coarser resolution images. The method proposed in this study effectively addresses the shortcomings of conventional FVC estimation methods, improves the accuracy of FVC monitoring in soil erosion areas, and serves as a reference for large-scale ecological environment monitoring using UAV technology.

Key message Evaluated tower, mast, crane, and UAV methods for forest vertical gap fraction, LAI, and CI measurements in different seasons. UAV is promising for forest vertical structural profiling. The vertical distribution of canopy structural parameters, such as canopy gap fraction, leaf area index (LAI) and clumping index (CI), is important for understanding the forest structural and functional properties. However, vertically distributed canopy structural data are rare, and current methods are either inefficient or costly for obtaining sufficient amounts of such data. This study conducted a series of field campaigns to obtain forest vertical structural measurements at two temperate forest sites in northern China from 2020 to 2023. Four different measurement systems were compared: (1) flux towers with accessible platforms at different heights, (2) a portable and extensible sampling mast with a digital hemispherical photography (DHP) camera attached on top, (3) a tower crane with a DHP camera fixed on the crane hook, and (4) an uncrewed aerial vehicle (UAV) with a DHP camera attached on top. The measured effective plant area index (PAI eff ) shows clearly seasonal variations at different heights. The CI remains relatively consistent at different heights, and the leaf-off value is approximately 0.1−0.2 higher than the leaf-on one. The flux tower method can be used for vertical profile measurement at a fixed location, whereas the portable mast is suitable for lower-level (< 15 m) measurement. Crane measurement requires an established facility and is useful for local measurement around the crane. UAV with an attached DHP provides a promising method for monitoring vertical structural parameters. The vertical structural profiles obtained in this study can be used in various modeling and validation studies.

Analyzing video data from an uncrewed aerial vehicle (UAV) of two short-lived dome building events at Anak Krakatau volcano (Indonesia), we determine vertical and horizontal movements of the dome surface prior to explosions, as well as initial eruption velocities and mass eruption rates via automated feature tracking and other photogrammetric methods. Initial eruption velocities and mass eruption rates are estimated as a proxy for eruptive strength. Eruptive strength is found to correlate with deformation magnitude, i.e., larger pre-explosion surface displacements are followed by both higher initial eruption velocities and mass fluxes. In accord with other studies, our observations can be explained by an overpressure underneath the dome’s surface. We assume that the dome seals the underlying vent efficiently, meaning that pre-explosion pressure build-up controls both deformation magnitude and eruptive strength. We support this assumption by a simple numerical model indicating that pre-explosion pressure increases between 8 and 16 MPa. The model further reveals that the two events vary significantly with respect to the importance of lateral visco-elastic flow for pressurization and deformation. The video sequences also show considerable variations in the gas release and associated deformation characteristics. Both constant and accelerating deformation is observed. Our case study demonstrates that photogrammetric methods are suitable to provide quantitative constraints on both effusive and explosive activity. Future work can build on our or similar approaches to develop automated monitoring strategies that would enable the observation and analysis of volcanic activity in near real time during a volcanic crisis.

Water quality monitoring of medium-sized inland water is important for water environment protection given the large number of small-to-medium size water bodies in China. A case study was conducted on Yuandang Lake in the Yangtze Delta region, with a surface area of 13 km2. This study proposed utilising a multispectral uncrewed aerial vehicle (UAV) to collect large-scale data and retrieve multiple water quality parameters using machine learning algorithms. An alternate processing method is proposed to process large and repetitive lake surface images for mapping the water quality data to the image. Machine learning regression methods (Random Forest, Gradient Boosting, Backpropagation Neural Network, and Convolutional Neural Network) were used to construct separate water quality inversion models for ten water parameters. The results showed that several water quality parameters (CODMn, temperature, pH, DO, and NC) can be retrieved with reasonable accuracy (R2 = 0.77, 0.75, 0.73, 0.67, and 0.64, respectively), although others (NH3-N, BGA, TP, Turbidity, and Chl-a) have a determination coefficient (R2) less than 0.6. This work demonstrated the tremendous potential of employing multispectral data in conjunction with machine learning algorithms to retrieve multiple water quality parameters for monitoring medium-sized bodies of water.