The proliferation of high-spatial-resolution remote sensing data is transforming forest attribute estimation, replacing traditional manual approaches with deep learning-based Individual Tree Crown Delineation (ITCD). Nevertheless, accurate ITCD boundary extraction from aerial RGB imagery faces persistent challenges: boundary ambiguity from complex crown occlusion in mixed forests, scarcity of high-quality annotations, and computational limitations of existing methods in dense forests. The latter manifests particularly in overlapping crown scenarios through constrained receptive fields, leading to substantial parameter requirements, computational inefficiency, and compromised accuracy. To overcome these limitations, we propose CrownViM, a novel architecture based on a bidirectional State Space Model (SSM). The framework integrates a linear-complexity Context Clustering Vision Mamba (CCViM) encoder for efficient global context modeling and employs a MaskFormer decoder for precise boundary prediction. We further introduce a partial-supervision loss function to reduce dependence on exhaustively annotated crown masks. Evaluations on OAM-TCD and the single-tree segmentation dataset (SSD) show CrownViM achieves significant segmentation accuracy improvements while maintaining a lightweight profile (39.6 M parameters). It substantially outperforms Convolutional Neural Network (CNN), Vision Transformer (ViT), and hybrid-based baselines when processing overlapping crowns and structurally complex scenes. As the first implementation of state space models in ITCD, CrownViM effectively addresses core limitations in global context capture, computational efficiency, and boundary definition. Our efficient architecture and sparse-annotation loss strategy enable high-accuracy, robust individual tree mapping, advancing tools for large-scale forest monitoring and accurate carbon stock quantification.

In photovoltaic remote sensing image segmentation tasks, fully supervised methods can achieve high accuracy. However, the high cost of pixel-level annotation significantly limits their scalability in large-scale scenarios. To overcome this annotation bottleneck, this paper proposes a zero-shot cross-modal segmentation framework based on the visual-language pre-trained foundation model (CLIP). This approach harnesses CLIP’s cross-modal knowledge transfer capabilities to achieve precise extraction of photovoltaic targets without requiring any downstream training. This paper first introduces the Layer-wise Augmented Residual Attention (LARA) mechanism to enhance fine-grained detail representation in the feature space. Subsequently, a Cross-modal Semantic Attribution Module (CMSA) is designed to generate precise activation maps by leveraging image-text alignment gradient information. Finally, the Confidence-Aware Refinement Strategy (CARS) replaces the conventional training-based denoising process, directly producing high-quality binary segmentation masks through adaptive thresholding. Comparative experiments were conducted to evaluate the proposed method against various baselines using several public datasets with varying resolutions in Jiangsu Province including Unmanned Aerial Vehicles imagery, Beijing-2, Gaofen-2, and a self-created Sentinel-2 imagery covering multiple countries. Notably, the proposed method achieved an IoU of 70.3% on the Gaofen-2 PV03 dataset with a spatial resolution of approximately 0.3 m and 50.8% on the self-created Sentinel-2 PV_Sentinel-2 dataset with a spatial resolution of 10 m. Experimental results demonstrate that our proposed approach maintains excellent cross-domain generalisation capabilities while reducing annotation costs, thereby providing an efficient and viable technical pathway for the automated monitoring of large-scale photovoltaic facilities.

High-resolution remote sensing technologies have increasingly become a versatile and cost-effective method for monitoring diverse ecosystems and enhancing vegetative species classification. This study utilizes two remote sensing approaches to classify the dominant, canopy-contributing tree species in a Bottomland Hardwood Forest (BHF) in Northeast Louisiana, evaluating the effectiveness of consumer-grade Unmanned Aerial Systems (UAS) and advanced machine learning techniques. High-resolution RGB aerial imagery was collected via UAS and processed using photogrammetric techniques to generate orthomosaics and texture features. Two classification methods were used: a U-Net Convolutional Neural Network (CNN) and a Random Forest classifier through an Object-based Image Analysis (OBIA) segmentation approach. The CNN approach demonstrated higher accuracy over OBIA in the classification of nine dominant tree species, particularly for less dominant species, with Quercus achieving the highest precision at 83.3%. This study highlights the potential use of UAS imagery and machine learning in forest management for species inventory and ecological monitoring. Results of this study indicate that while RGB imagery can effectively classify tree species, further integration with multi-temporal datasets and more powerful sensor combinations may enhance classification accuracy and enable practical applicability in environmental monitoring. The findings support the growing role of UAS remote sensing in ecological research and forest conservation efforts.

Vision-based unmanned aerial vehicles (UAVs) have been widely studied and applied in aerial monitoring tasks; however, detecting small objects in UAV imagery remains challenging due to limited visual features, significant scale variations, dense distributions, and complex background interference. In real-world UAV scenarios, small objects often occupy only a few pixels and are easily obscured by cluttered backgrounds, which complicates stable and accurate detection. To address these issues, this study proposes MSCM-YOLO, a UAV-oriented lightweight detection framework based on YOLOv11. The framework integrates four key innovations: (1) a dedicated P2 detection head to preserve high-resolution features for extremely small and dense targets; (2) a lightweight backbone enhanced with Mobile Bottleneck Convolution (MBConv) to improve feature extraction for visually weak objects; (3) a Scale-Adaptive Attention Fusion (SAF) mechanism with a Channel-Adaptive Projection (CAP) module to effectively integrate multi-scale spatial and semantic features under large object-size variations; and (4) a Multi-Domain Feature Attention Fusion (MDFAF) module to enhance target-background discrimination in complex UAV scenes. Experiments on the VisDrone2019 dataset show that MSCM-YOLO achieves mAP50 and mAP50:95 scores of 44.41% and 27.13%, respectively, outperforming the YOLOv11 baseline by 10.77 and 7.22 percentage points. Notably, the proposed framework achieves this significant performance improvement while maintaining a balanced computational profile suitable for UAV deployment. Additional validation on the UAVDT, DIOR, and AI-TOD datasets confirms consistent improvements in mAP50, demonstrating the robustness and generalization ability of the proposed method. Overall, MSCM-YOLO provides an effective and practical solution for accurate small object detection in aerial monitoring applications.

Greenhouse agriculture plays a vital role in sustainable food production but also entails challenges such as high energy consumption and significant carbon emissions. To address these issues and evaluate the feasibility of adopting alternative energy sources such as green hydrogen, it is essential to precisely understand the spatial structure of agricultural facilities and accurately predict their energy demands. However, effectively processing real-world datasets—characterized by complex aerial imagery and severe class imbalance—remains a technical challenge. This study proposes a deep learning-based framework for accurately detecting and segmenting greenhouses using high-resolution aerial images. To improve object detection performance, a class-weighted loss function and a class-aware sampling strategy were integrated into the You Only Look Once 8 (YOLOv8) model to mitigate the effects of class imbalance. The proposed model achieved an overall mean average precision (mAP)@0.5:0.95 of 0.566, with precision increasing to 0.881 and recall improving to 0.822, demonstrating balanced and robust performance across classes. Additionally, the model was combined with the Segment Anything Model (SAM) to enhance segmentation precision, and its performance was compared against Open-World Localization Vision Transformer (OWL-ViT) + SAM and YOLOv8-only segmentation approaches. Experimental results show that the YOLOv8 + SAM (Fusion) configuration achieved the highest Intersection over Union (IoU) of 0.7588 and Dice coefficient (Dice) of 0.8578, demonstrating superior boundary accuracy and mask consistency compared to other methods. The joint application of class weighting and sampling improved recall for minority classes such as greenhouses, while SAM-based segmentation enhanced boundary fidelity and shape preservation. Based on the segmented areas, greenhouse surface areas were calculated, and a conservative energy consumption benchmark was applied to estimate annual energy demand. This research presents a practical baseline for evaluating the potential of renewable energy integration in agriculture and is expected to contribute to future strategies for achieving carbon neutrality and green hydrogen utilization in the agricultural sector.

In autonomous exploration tasks in unstructured terrain, exploration efficiency and map topology quality have been a difficult problem to balance. Among the current autonomous exploration methods, geometry-based exploration methods only focus on exploration efficiency but not map quality, which not only leads to frequent backtracking by the robot, but also tends to ignore non-geometric risks such as negative obstacles. To address this pain point, we propose the Structure-Aware Topology Exploration framework. Unlike pure geometric exploration, we utilize U-Net to semantically analyze the unmanned aerial vehicle aerial images, and force the robot’s path to be anchored to the geometric axis of the safe area through the Semantic Seeded Voronoi mechanism. To avoid map redundancy leading to backtracking, we directly introduce topological sparsity constraints in the decision function to realize online structural pruning during exploration. Simulation experiments based on real-world aerial imagery demonstrate that the proposed framework effectively overcomes the late-stage exploration plateau: compared with purely geometric baselines (Rapidly exploring Random Tree and Frontier), it reduces average path length to 278.4 m (45% reduction) and improves exploration efficiency by 80%; compared with the semantic frontier-based baseline, it achieves 28.6% higher efficiency and 13% shorter path length, maximizing information gain per unit travel distance.

MMP-Mapper: Multi-modal priors enhancing vectorized HD road map construction from aerial imagery

Small object detection in aerial imagery is a challenging task due to the minimal pixel information in dense clutter, scale variation, and complex backgrounds. YOLOv9 has demonstrated the effectiveness of Programmable Gradient Information (PGI) in mitigating feature degradation. However, its fully convolutional architecture lacks the capability for global context modeling, which is critical for resolving ambiguities in small targets. To address these limitations, we propose SF-YOLOv9, a hybrid architecture that enhances YOLOv9c by improving the backbone through the integration of a novel PGI-Aware Swin Fusion Block (Transformer-GELAN) at its final stage. This module effectively preserves high-resolution local features while injecting long-range global context through Swin Transformer-based fusion. It results in richer and more discriminative semantic representations. We introduce a Dual-Path Spatial and Channel Attention Module (DSCAM) into the main detection head and the reversible auxiliary branches of PGI. By refining attention across all supervisory signals, DSCAM significantly improves gradient flow and feature fidelity during PGI training, reducing missed detections and false positives. We evaluate SF-YOLOv9 on VisDrone and NWPU-VHR-10 datasets to demonstrate the effectiveness of SF-YOLOv9. It outperformed the baseline models, achieving 49.1% mAP@0.50 on VisDrone and 98.3% mAP@0.50 on NWPU VHR-10 in small-object detection.

Freq-DETR: Frequency-aware transformer for real-time small object detection in unmanned aerial vehicle imagery

Water stress significantly threatens sugarcane production, particularly among smallholder farmers in South Africa, where spatially explicit assessments remain limited. This study aimed to improve the quantification of crop water stress by developing a machine learning (ML) model to predict the Normalised Difference Water Index (NDWI), a proxy for vegetation water content. An ML approach was adopted to capture complex, non-linear relationships between structural vegetation indices (SVIs) and NDWI. Sentinel-2 satellite data and UAV-acquired multispectral imagery were integrated, with the model trained using satellite-derived SVIs and NDWI, and then applied to UAV-derived SVIs to predict NDWI. The model achieved high predictive accuracy (R² = 0.95, RMSE = 0.03, MAE = 0.02) and effectively captured temporal variations in sugarcane water status, including post-rainfall stress recovery and increased water retention during early maturation—aligning with changes in leaf area index (LAI), chlorophyll content (CC), and Total Soil Water Profile (TSWP). NDWI also showed a positive correlation with actual evapotranspiration (ET a ; R² = 0.60) and a negative correlation with the Water Deficit Index (WDI; R² = 0.62), suggesting its potential to reflect crop water status under certain conditions. When interpreted in conjunction with in situ measurements of precipitation, TSWP, and WDI, the predicted NDWI provides valuable insights into crop water dynamics. This approach demonstrates the potential of ML-driven NDWI estimation to support site-specific irrigation scheduling, enhance resource use efficiency, and promote sustainable sugarcane cultivation. The findings contribute to climate-resilient water management practices tailored to the needs of smallholder systems in water-scarce regions. • Machine learning predicts multispectral UAV-derived NDWI using Sentinel-2 vegetation indices. • Predicted NDWI aligns with trends in soil water status, evapotranspiration and water deficit dynamics. • NDWI correlates positively with ET a and negatively with WDI, reflecting crop water stress levels. • These relationships capture shifts in crop water dynamics under varying environmental and meteorological conditions. • Model outputs can inform timely, site-specific irrigation strategies in rainfed sugarcane production systems.

DRM-YOLO: A YOLOv11-based structural optimization method for small object detection in UAV aerial imagery

A class-added rail transit infrastructure object detection method using UAV aerial imagery based on self-ensemble masks and frequency-spatial calibration

Abstract Litter pollution remains a pressing environmental issue, motivating the search for automated monitoring methods that can scale effectively. The problem is heightened by the difficulty of detecting small and diverse litter objects across wide areas, prompting interest in Unmanned Aerial Vehicles (UAVs) and deep learning as viable solutions. We conducted a systematic literature review by initially using Google Scholar, and then iteratively expanding our search through bibliographic references to identify relevant studies and datasets. In this review, we: synthesize the applicability of nine publicly available litter datasets; compile and analyze computer vision integrations with Litter Management Systems, with a particular emphasis towards UAV-based solutions; and review relevant literature addressing this issue; among others. Our analysis includes four UAV-based datasets (BDW, UAVVaste, HAIDA, SODA) and five non-UAV datasets (TrashNet, TACO, MJU-Waste, PlastOPol, ZeroWaste), examining dataset characteristics, preprocessing techniques, model architectures, and evaluation metrics across studies. Our synthesis of the literature highlights the varied approaches different studies undertake, reflecting the complexity of the task and the absence of standardised protocols. We conclude by discussing priorities for future work, notably the need for more publicly available in-the-wild UAV-acquired datasets and the potential of newer model architectures to address current limitations in automated litter detection.

Caribou populations across the Arctic have declined markedly in recent decades, motivating scalable, consistent, and accurate monitoring approaches to guide evidence-based conservation actions and policy decisions. By providing broad coverage through high-resolution imagery, aerial surveys offer a practical means to monitor wildlife across vast and remote Arctic regions. Manual interpretation from this imagery is labor-intensive and error-prone, underscoring the need for automatic and reliable detection across varying scenes. Yet, such automatic detection is particularly challenging due to severe background heterogeneity, dominant empty terrain (class imbalance), small or occluded targets, and wide variation in density and scale. To make the detection model (HerdNet) more robust to these challenges, a weakly supervised patch-level pretraining based on a detection network’s architecture is proposed. The detection dataset includes five caribou herds distributed across Alaska. By learning from empty vs. non-empty labels in this dataset involving heterogeneous Arctic scenes, the approach produces early weakly supervised knowledge for enhanced detection compared to HerdNet, which is initialized from generic weights. Accordingly, the patch-based pretrain network attained high accuracy on multi-herd imagery (2017) and on an independent year’s (2019) test sets (F1: 93.7%/92.6%, respectively), enabling reliable mapping of regions containing animals to facilitate manual counting on large aerial imagery. Transferred to detection, initialization from weakly supervised pretraining yielded consistent gains over ImageNet weights on both positive patches (F1: 92.6%/93.5% vs. 89.3%/88.6%), and full image counting (F1: 95.5%/93.3% vs. 91.5%/90.4%). Remaining limitations are dominated by false positives from animal-like background clutter, and false negatives related to low animal density occlusions. Overall, pretraining on coarse labels prior to detection makes it possible to rely on weakly-supervised pretrained weights even when labeled data are limited, achieving results comparable to generic-weight initialization.

Unmanned Aerial Vehicle (UAV) remote sensing has become essential for real-time earth observation applications, including precision agriculture, traffic monitoring, and disaster response. However, small-target detection in UAV aerial imagery still faces critical challenges: extreme scale variation due to variable flight altitudes, background interference from complex terrain, and insufficient pixel information for tiny objects. To address these issues, this work proposes FKIFM-DETR, a real-time transformer-based detection framework leveraging multi-domain information fusion. First, a Spatial-Frequency Fusion Module (SFM) is designed to integrate spatial and frequency-domain features for capturing fine-grained target details while suppressing background noise; second, a High–Low Frequency Block (HL-Block) is introduced to separately process high-frequency local details and low-frequency global context, balancing detail retention and semantic awareness; finally, a Channel Feature Recalibration-Enhanced Feature Pyramid Network (SPCR-FPN) is employed to strengthen the interaction between shallow spatial features and deep semantic features. On the VisDrone2019 dataset, FKIFM-DETR achieves 6.3% and 5.3% improvements in mAP@0.5 and mAP@0.5:0.95 over the RT-DETR baseline, respectively; evaluations on TinyPerson and HIT-UAV datasets further demonstrate its cross-scenario applicability. These results demonstrate the potential of FKIFM-DETR for practical UAV remote sensing applications such as crowd surveillance, vehicle tracking, and emergency rescue.

Recent advances in remote sensing have increasingly emphasized multi-view vision, which integrates complementary viewpoints to deliver more complete scene understanding and effectively alleviate occlusion and limited fields of view in crowded environments. In particular, aerial imagery captured by drones provides holistic scene coverage, whereas ground-level cameras offer precise and fine-grained object details. Despite these advantages, large-scale multi-view datasets that jointly incorporate aerial and ground-level perspectives remain scarce, largely due to the practical difficulties of coordinating paired aerial and ground platforms. To overcome this challenge, we develop a ground–aerial camera system that emulates drone viewpoints and, based on this system, construct a large-scale synthetic dataset for aerial–ground multi-view person association. Leveraging this dataset, we propose a novel graph-constrained framework that enforces robust and globally consistent associations across aerial and ground views. Additionally, we introduce an aerial-view-guided people-number estimation module to provide a scene-level constraint for identity association. Extensive experimental results demonstrate that our method consistently outperforms state-of-the-art baselines in multi-view labeling across varying crowd densities.

Aerial Imagery as a Window into Local Wisdom for Community Development: A Case Study from Rural Paraguay

Object detection in UAV aerial imagery plays a pivotal role across a wide spectrum of applications. However, existing detection models continue to face significant challenges stemming from small object scales, dense spatial distributions, and highly complex backgrounds. To address these challenges, this paper proposes a novel dual-backbone network model named WCDB-YOLO. The core innovation of this work lies in introducing a “target-context decoupled perception” paradigm, which utilizes two structurally complementary backbone networks to separately process local object features and global background information: one backbone focuses on extracting fine-grained local features of objects, while the other innovatively incorporates a wavelet convolution module to efficiently model the global contextual semantics of complex scenes with minimal computational cost by constructing a large receptive field. To further enhance the scale adaptability for small objects, a Dilation-wise Residual (DWR) module is designed, which employs parallel convolutional branches with different dilation rates to achieve dynamic adaptation to multi-scale small object features. Additionally, the model optimizes the feature pyramid structure by integrating high-resolution P2/4 features into the detection head, significantly improving the localization accuracy of tiny objects. Experimental results on the VisDrone dataset show that the proposed method achieves an 8.4% improvement in mAP50 over the baseline YOLOv11s model and outperforms current state-of-the-art (SOTA) approaches. This work presents a highly accurate and robust solution for small object detection from UAV platforms in complex environments.

UAV imagery-based semantic segmentation is crucial for mapping tropical agricultural areas such as oil palm plantations. The main challenges are overlapping vegetation objects, unclear boundaries, and spectral similarities between classes, which reduce the accuracy of conventional models. This study proposes a modified U-Net architecture with a VGG-19 backbone, achieved through hyperparameter tuning (M7) and the integration of residual blocks (M8), to enhance multi-class segmentation performance. Experiments were conducted on aerial imagery with two resolutions (512×512 and 256×256) using four-class and three-class scenarios. The results show that M7 and M8 consistently outperform the baseline model (M2) in terms of accuracy, precision, recall, and average Intersection over Union (IoU). In the 512x512 four-class scenario, M8 achieved the highest accuracy (87.40%), precision (88.32%), recall (86.32%), and MIoU (0.132). M7 reached similar accuracy (>86%) but trained significantly faster than the baseline. In the 256x256 scenario, M8 maintained strong performance with 86.44% accuracy and 0.302 MIoU. For the three-class experiment, M8 reached a top MIoU of 0.178. Accuracy, precision, and recall were all above 87%, showing improved recognition of minority classes such as waterways. Confusion matrix analysis confirmed that M8 provided more balanced class predictions. It also reduced false negatives for oil palm vegetation. M7 showed slight fluctuations, suggesting possible overfitting. These findings support M8 as a robust solution for UAV-based oil palm mapping and large-scale monitoring.

ABSTRACT Neolithic long barrows are among the earliest monumental structures in Europe, yet in many parts of Central Europe their surface expression has been largely erased by long‐term agricultural activity. This study evaluates the potential of integrated remote sensing approaches for identifying and contextualizing long barrows and associated archaeological features in the intensively cultivated landscape of north‐western Bohemia, Czech Republic. The analysis focuses on two long barrows near Mount Říp (Dušníky 1 and Nížebohy) and combines oblique aerial photography, magnetic survey and airborne laser scanning (ALS). The results demonstrate that even heavily levelled long barrows remain detectable through the complementary use of non‐invasive methods. ALS confirms the presence of subdued barrow mantles, while aerial imagery and magnetometry provide key information on monument layout, internal structure and associated burial features. In addition, the remote sensing data reveal further funerary monuments of later prehistoric periods, indicating long‐term continuity of burial practices and the development of enduring ritual landscapes. Settlement activities are spatially restricted and clearly separated from funerary zones. This study highlights the value of multi‐method remote sensing for reconstructing prehistoric burial landscapes in intensively cultivated regions.