Unmanned Aerial Vehicle Images Research Articles

Weed Detection Algorithms in Rice Fields Based on Improved YOLOv10n

Weeds in paddy fields compete with rice for nutrients and cause pests and diseases, greatly affecting rice yield. Accurate weed detection is vital for implementing variable spraying with unmanned aerial vehicles (UAV) for weed control. Therefore, this paper presents an improved weed detection algorithm, YOLOv10n-FCDS (YOLOv10n with FasterNet, CGBlock, Dysample, and Structure of Lightweight Detection Head), using UAV images of Sagittaria trifolia in rice fields as the research object, to address challenges like the detection of small targets, obscured weeds and weeds similar to rice. We enhanced the YOLOv10n model by incorporating FasterNet as the backbone for better small target detection. CGBlock replaced standard convolution and SCDown modules to improve the detection ability of obscured weeds, while DySample enhanced discrimination between weeds and rice. Additionally, we proposed a lightweight detection head based on shared convolution and scale scaling, maintaining accuracy while reducing model parameters. Ablation studies revealed that YOLOv10n-FCDS achieved a 2.6% increase in mean average precision at intersection over union 50% for weed detection, reaching 87.4%. The model also improved small target detection (increasing mAP50 by 2.5%), obscured weed detection (increasing mAP50 by 2.8%), and similar weed detection (increasing mAP50 by 3.0%). In conclusion, YOLOv10n-FCDS enables effective weed detection, supporting variable spraying applications by UAVs in rice fields.

Pit slope unmanned aerial vehicle image matching based on hybrid scale space and boosted efficient binary local image descriptor

Pit slope unmanned aerial vehicle image matching based on hybrid scale space and boosted efficient binary local image descriptor

Mamba-UAV-SegNet: A Multi-Scale Adaptive Feature Fusion Network for Real-Time Semantic Segmentation of UAV Aerial Imagery

Accurate semantic segmentation of high-resolution images captured by unmanned aerial vehicles (UAVs) is crucial for applications in environmental monitoring, urban planning, and precision agriculture. However, challenges such as class imbalance, small-object detection, and intricate boundary details complicate the analysis of UAV imagery. To address these issues, we propose Mamba-UAV-SegNet, a novel real-time semantic segmentation network specifically designed for UAV images. The network integrates a Multi-Head Mamba Block (MH-Mamba Block) for enhanced multi-scale feature representation, an Adaptive Boundary Enhancement Fusion Module (ABEFM) for improved boundary-aware feature fusion, and an edge-detail auxiliary training branch to capture fine-grained details. The practical utility of our method is demonstrated through its application to farmland segmentation. Extensive experiments on the UAV-City, VDD, and UAVid datasets show that our model outperforms state-of-the-art methods, achieving mean Intersection over Union (mIoU) scores of 71.2%, 77.5%, and 69.3%, respectively. Ablation studies confirm the effectiveness of each component and their combined contributions to overall performance. The proposed method balances segmentation accuracy and computational efficiency, maintaining real-time inference speeds suitable for practical UAV applications.

Detection of Rice Leaf Folder in Paddy Fields Based on Unmanned Aerial Vehicle-Based Hyperspectral Images

Pest infestations significantly impact rice production and threaten food security. Remote sensing offers a vital tool for the non-destructive, rapid detection of rice pests. Existing studies often focus on laboratory conditions at the leaf level, limiting their applicability for precise pesticide application. Therefore, this study aimed to develop a method for detecting rice pests (rice leaf folders) in paddy fields based on unmanned aerial vehicle (UAV) hyperspectral data. Firstly, a UAV imaging system collected hyperspectral images of rice plants in both the jointing and heading stages. A total of 222 field plots for investigating rice leaf folders was established during these two periods. Secondly, 23 vegetation indices were calculated as candidates for identifying rice pests. Then, hyperspectral data and field investigation data from the jointing stage were used to construct a machine learning (extreme gradient boosting, XGBoost) algorithm for detecting rice pests. The results showed that the XGBoost model exhibited the best performance when eight vegetation indices were considered as the selected input features for model construction: the Red-edge Normalized Difference Vegetation Index (red-edge NDVI), Structure Insensitive Pigment Index (SIPI), Enhanced Vegetation Index (EVI), Atmospherically Resistant Vegetation Index (ARVI), Soil-Adjusted Vegetation Index (SAVI), Red-edge Chlorophyll Index (CIred-edge), Pigment-Specific Simple Ratio680 (PSSR680), and Carotenoid Reflectance Index700 (CPI700). The training and testing accuracies reached 87.46% and 86%, respectively. Furthermore, the heading stage application confirmed the model’s feasibility. Thus, the XGBoost model with input features of eight vegetation indices provides an effective and reliable method for detecting rice leaf folders, supporting real-time, precise pesticide use in rice cultivation.

PYRAMID CHANNEL-BASED FEATURE ATTENTION NETWORK WITH ENSEMBLE LEARNING-BASED UAV IMAGE CLASSIFICATION ON IoT-ASSISTED REMOTE SENSING ENVIRONMENT

Advances in Unmanned Aerial Vehicles (UAVs), otherwise recognized as drones, have tremendous promise in improving the wide-ranging applications of the Internet of Things (IoT). UAV image classification using deep learning (DL) is an amalgamation to modernize data analysis, collection, and decision-making in a variety of sectors. IoT devices collect information in real time, while remote sensing captures data afar without direct contact. UAVs equipped with sensors offer high-quality images for classification tasks. DL techniques, especially the convolutional neural networks (CNNs), analyze data streams, extracting complicated features for the accurate classification of objects or environmental features. This synergy enables applications including urban planning and precision agriculture, fostering smarter disaster response, decision support systems, and efficient resource management. This paper introduces a novel Pyramid Channel-based Feature Attention Network with an Ensemble Learning-based UAV Image Classification (PCFAN-ELUAVIC) technique in an assisted remote sensing environment. The PCFAN-ELUAVIC technique begins with the contrast enhancement of the UAV images using the CLAHE technique. Following that, the feature vectors are derived by the use of the PCFAN model. Meanwhile, the hyperparameter tuning procedure is executed by the inclusion of a vortex search algorithm (VSA). For image classification, the PCFAN-ELUAVIC technique comprises an ensemble of three classifiers like long short-term memory (LSTM), graph convolutional networks (GCNs), and Hermite neural network (HNN). To exhibit the improved detection results of the PCFAN-ELUAVIC system, an extensive range of experiments are carried out. The experimental values confirmed the enhanced performance of the PCFAN-ELUAVIC model when compared to other techniques.

CDANet: a small object detection model for unmanned aerial vehicle images based on cross-layer guidance and skip dilated fusion

CDANet: a small object detection model for unmanned aerial vehicle images based on cross-layer guidance and skip dilated fusion

Using UAV-based multispectral and RGB imagery to monitor above-ground biomass of oat-based diversified cropping

Timely access to crop above-ground biomass (AGB) information is crucial for estimating crop yields and managing water and fertilizer efficiently. Unmanned aerial vehicle (UAV) imagery offers promising avenues for AGB estimation due to its high efficiency and flexibility. However, the accuracy of these estimations can be influenced by various factors, including crop growth stages, the spectral resolution of UAV sensors, and flight altitudes. These factors need thorough investigation, especially in diversified cropping systems where crop diversity and growth stages interplay complexly, challenging the accuracy of AGB estimation. This study aims to estimate AGB of oats planted under different agricultural regimes—monoculture, crop rotation, and intercropping—at various growth stages (jointing, flowering, and grain-filling) and across all stages combined, using multispectral and RGB UAV images collected at different flight altitudes (25 m, 50 m, and 100 m). Three feature selection methods—maximal information coefficient (MIC), least absolute shrinkage and selection operator (LAS), and recursive feature elimination (RFE)—were tested. Four machine learning models—ridge regression (RR), multilayer perceptron (MLP), light gradient boosting machine (LGBM), and extreme gradient boosting (XGBoost)—were used for estimating AGB. Each feature selection method was combined with each machine learning model (e.g., MIC-RR, MIC-MLP, MIC-LGBM, MIC-XGBoost, LAS-RR) to evaluate their performance. Results revealed that the highest accuracy in AGB estimation was achieved with images acquired at a flight altitude of 25 m. The RFE-MLP model demonstrated superior results during the jointing stage (R² = 0.84, root mean squared error (RMSE) = 217.45 kg/ha, root mean squared logarithmic error (RMSLE) = 0.16, mean absolute percentage error (MAPE) = 4.15 %), the LAS-RR model excelled in the flowering stage (R² = 0.85, RMSE = 263.03 kg/ha, RMSLE = 0.05, MAPE = 14.44 %), and the RFE-XGBoost model was most effective during the grain-filling stage (R² = 0.68, RMSE = 865.03 kg/ha, RMSLE = 0.12, MAPE = 8.88 %). For cross-stage modelling, the RFE-MLP achieved remarkable results (R² = 0.93, RMSE = 680.44 kg/ha, RMSLE = 0.16, MAPE = 12.12 %). This study demonstrates the efficacy of combining feature selection methods with machine learning algorithms to enhance the accuracy of oat AGB estimations. The involvement of multiple cropping patterns enhances the generalizability of our findings, facilitating real-time and rapid monitoring of crop growth in future diversified cropping systems.

Development of methodology for calculating flooded area and flood volume in small urban areas based on unmanned aerial vehicle images

Climate change has intensified flooding and increased localized torrential rainfalls, leading to disasters such as landslides, infrastructure collapse, and urban floods. The extent and accuracy of flood damage information significantly impact recovery processes. While previous studies primarily utilized satellite and aerial imagery for broad flood assessments, they often lacked the precision needed for accurate damage analysis. This study addresses the gap between rapid assessment needs and precise damage quantification in flood inundation analysis. This research introduces a novel image-based investigation approach to enhance the speed and accuracy of flood inundation assessment. By leveraging unmanned aerial vehicles (UAVs) and image-based spatial data technology, aerial images of flooded areas are rapidly captured to construct three-dimensional disaster site terrain information. The proposed methodology employs advanced techniques in aerial photography, image processing, and geographic analysis to quantitatively analyze flood inundation scale using only aerial images and geographic information systems (GIS). The research yielded a calculated flood inundation area of 3,847.36 m2 and a flood volume of 13,895.13 m3. This methodology complements existing flood inundation assessment techniques and has the potential to significantly improve disaster management efforts by providing rapid, accurate, and actionable data for decision-makers.

Machine Learning Based Inversion of Water Quality Parameters in Typical Reach of Rural Wetland by Unmanned Aerial Vehicle Images

Rural wetlands are complex landscapes where rivers, croplands, and villages coexist, making water quality monitoring crucial for the well-being of nearby residents. UAV-based imagery has proven effective in capturing detailed features of water bodies, making it a popular tool for water quality assessments. However, few studies have specifically focused on drone-based water quality monitoring in rural wetlands and their seasonal variations. In this study, Xiangfudang Rural Wetland Park, Jiaxin City, Zhejiang Province, China, was taken as the study area to evaluate water quality parameters, including total nitrogen (TN), total phosphors (TP), chemical oxygen demand (COD), and turbidity degree (TUB). We assessed these parameters across summer and winter seasons using UAV multispectral imagery and field sample data. Four machine learning algorithms were evaluated and compared for the inversion of the water quality parameters, based on the situ sample survey data and UAV multispectral images. The results show that ANN algorithm yielded the best results for estimating TN, COD, and TUB, with validation R2 of 0.78, 0.76, and 0.57, respectively; CatBoost performed best in TP estimation, with validation R2 and RMSE values of 0.72 and 0.05 mg/L. Based on spatial estimation results, the average COD concentration in the water body was 16.05 ± 9.87 mg/L in summer, higher than it was in winter (13.02 ± 8.22 mg/L). Additionally, mean TUB values were 18.39 Nephelometric Turbidity Units (NTU) in summer and 20.03 NTU in winter. This study demonstrates the novelty and effectiveness of using UAV multispectral imagery for water quality monitoring in rural wetlands, providing critical insights into seasonal water quality variations in these areas.

Methods for Extracting Fractional Vegetation Cover from Differentiated Scenarios Based on Unmanned Aerial Vehicle Imagery

Fractional vegetation cover (FVC) plays a key role in ecological and environmental status assessment because it directly reflects the extent of vegetation cover and its status, yet vegetation is an important component of ecosystems. FVC estimation methods have evolved from traditional manual interpretation to advanced remote sensing technologies, such as satellite data analysis and unmanned aerial vehicle (UAV) image processing. Extraction methods based on high-resolution UAV data are being increasingly studied in the fields of ecology and remote sensing. However, research on UAV-based FVC extraction against the backdrop of the high soil reflectance in arid regions remains scarce. In this paper, based on 12 UAV visible light images in differentiated scenarios in the Ebinur Lake basin, Xinjiang, China, various methods are used for high-precision FVC estimation: Otsu’s thresholding method combined with 12 Visible Vegetation Indices (abbreviated as Otsu-VVIs) (excess green index, excess red index, excess red minus green index, normalized green–red difference index, normalized green–blue difference index, red–green ratio index, color index of vegetation extraction, visible-band-modified soil-adjusted vegetation index, excess green minus red index, modified green–red vegetation index, red–green–blue vegetation index, visible-band difference vegetation index), color space method (red, green, blue, hue, saturation, value, lightness, ‘a’ (Green–Red component), and ‘b’ (Blue–Yellow component)), linear mixing model (LMM), and two machine learning algorithms (a support vector machine and a neural network). The results show that the following methods exhibit high accuracy in FVC extraction across differentiated scenarios: Otsu–CIVE, color space method (‘a’: Green–Red component), LMM, and SVM (Accuracy > 0.75, Precision > 0.8, kappa coefficient > 0.6). Nonetheless, higher scene complexity and image entropy reduce the applicability of precise FVC extraction methods. This study facilitates accurate, efficient extraction of vegetation information in differentiated scenarios within arid and semiarid regions, providing key technical references for FVC estimation in similar arid areas.

Florida mangrove dieback on a decadal and centennial timescales

Hurricanes are considered some of the most devastating weather phenomena, causing deaths and destruction along the coast. Mangroves are perceived as natural defenses against coastal hazards. However, mangrove gaps, known as mangrove dieback, have occurred in southwestern Florida. This study implemented a spatial-temporal analysis based on satellite, lidar and unmanned aerial vehicle (UAV) images collected along the southwestern Florida coast between Tampa and Naples from 1994 to 2020 CE, and multi-proxy analyses on sediment cores sampled from two mangrove dieback sites to determine the cause(s) of mangrove death and the implications for the sustainability of these forests. The data indicated that out of the 86 identified dieback, 78 % arose between 2004 and 2017 CE, when one tropical storm and three hurricanes made landfall on the studied coast, suggesting hurricanes as the main driver for this mangrove degradation. Multi-proxy analysis indicated mangrove expansion since ~600 cal yr BP due to relative sea-level (RSL) rise. However, hurricanes caused losses in the mangrove area between 1930 and 1950 CE, and during the last two decades in the studied sites. Ground topographic data and aerophotogrammetry revealed circular dieback with permanent ponds (~30 cm depth) inhibiting mangrove recovery. The mangrove death occurred in <3 years, with a gradual dieback expansion due to positive feedback involving physicochemical processes in the sediments. In addition, recent hurricanes increased pre-existing dieback areas. Therefore, dieback sites are vulnerability hotspots for mangroves in the face of projected RSL rise and higher frequency of intense hurricane strikes. Mangrove dieback should be monitored on a temporal (seasonal - annual) and spatial (cm-m) resolution only possible through the combination of satellite and UAV data. Integrating remote sensing and stratigraphic analysis enabled us to identify mangrove diebacks on a decadal to centennial timescale, improving our comprehension of how catastrophic weather events affect these forests.

LightUAV-YOLO: a lightweight object detection model for unmanned aerial vehicle image

LightUAV-YOLO: a lightweight object detection model for unmanned aerial vehicle image

Improved Early-Stage Maize Row Detection Using Unmanned Aerial Vehicle Imagery

Monitoring row centerlines during early growth stages is essential for effective production management. However, detection becomes more challenging due to weed interference and crop row intersection in images. This study proposed an enhanced Region of Interest (ROI)-based approach for detecting early-stage maize rows. It integrated a modified green vegetation index with a dual-threshold algorithm for background segmentation. The median filtering algorithm was also selected to effectively remove most noise points. Next, an improved ROI-based feature point extraction method was used to eliminate residual noises and extract feature points. Finally, the least square method was employed to fit the row centerlines. The detection accuracy of the proposed method was evaluated using the unmanned aerial vehicle (UAV) image data set containing both regular and intersecting crop rows. The average detection accuracy of the proposed approach was between 0.456° and 0.789° (the angle between the fitted centerline and the expert line), depending on whether crop rows were regular/intersecting. Compared to the Hough Transform (HT) algorithm, the results demonstrated that the proposed method achieved higher accuracy and robustness in detecting regular and intersecting crop rows. The proposed method in this study is helpful for refined agricultural management such as fertilization and irrigation. Additionally, it can detect the missing-seedling regions and replenish seedings in time to increase crop yields.

Benchmarking of Individual Tree Segmentation Methods in Mediterranean Forest Based on Point Clouds from Unmanned Aerial Vehicle Imagery and Low-Density Airborne Laser Scanning

Three raster-based (RB) and one point cloud-based (PCB) algorithms were tested to segment individual Aleppo pine trees and extract their tree height (H) and crown diameter (CD) using two types of point clouds generated from two different techniques: (1) Low-Density (≈1.5 points/m2) Airborne Laser Scanning (LD-ALS) and (2) photogrammetry based on high-resolution unmanned aerial vehicle (UAV) images. Through intensive experiments, it was concluded that the tested RB algorithms performed best in the case of UAV point clouds (F1-score &gt; 80.57%, H Pearson’s r &gt; 0.97, and CD Pearson´s r &gt; 0.73), while the PCB algorithm yielded the best results when working with LD-ALS point clouds (F1-score = 89.51%, H Pearson´s r = 0.94, and CD Pearson´s r = 0.57). The best set of algorithm parameters was applied to all plots, i.e., it was not optimized for each plot, in order to develop an automatic pipeline for mapping large areas of Mediterranean forests. In this case, tree detection and height estimation showed good results for both UAV and LD-ALS (F1-score &gt; 85% and &gt;76%, and H Pearson´s r &gt; 0.96 and &gt;0.93, respectively). However, very poor results were found when estimating crown diameter (CD Pearson´s r around 0.20 for both approaches).

Evapotranspiration measurements in pasture, crops, and native Brazilian Cerrado based on UAV-borne multispectral sensor.

In Brazil, agriculture consumes most of the available freshwater, especially in the Cerrado biome, where the rain cycle is marked by long periods of drought. This study, conducted at the Brazilian Agricultural Research Corporation (Embrapa) Research Corporation unit in Santo Antônio de Goiás, Goiás, Brazil, estimated evapotranspiration (ET) in different crops and soil cover. Using multispectral unmanned aerial vehicle (UAV) images, Sentinel satellite data, weather station information, and towers employing the eddy covariance method, we applied the "Simple Algorithm for Evapotranspiration Retrieving" (SAFER) to calculate ET in common bean, pasture, and semideciduous seasonal forest areas. The results showed a good agreement between UAV and satellite data, with R2 = 0.84, also validated with flow towers by the eddy covariance method. UAV-based ET was observed to correspond well to tower (EC) during full vegetative development of beans but is underestimated at the beginning of planting and in the final periods of plant senescence, due to the influence of soil or straw cover. These findings contribute to a better understanding of water dynamics in the system and to enhancing sustainable agricultural practices. This method, adapted for multispectral aerial imaging, can be applied flexibly and on-demand, in different contexts and ground cover. The study highlights the importance of integrated agricultural practices for better management of water resources and preservation of the Cerrado in balance with cultivation areas.

Physics-Driven Image Dehazing from the Perspective of Unmanned Aerial Vehicles

Drone vision is widely used in change detection, disaster response, and military reconnaissance due to its wide field of view and flexibility. However, under haze and thin cloud conditions, image quality is usually degraded due to atmospheric scattering. This results in issues like color distortion, reduced contrast, and lower clarity, which negatively impact the performance of subsequent advanced visual tasks. To improve the quality of unmanned aerial vehicle (UAV) images, we propose a dehazing method based on calibration of the atmospheric scattering model. We designed two specialized neural network structures to estimate the two unknown parameters in the atmospheric scattering model: the atmospheric light intensity A and medium transmission t. However, calculation errors always occur in both processes for estimating the two unknown parameters. The error accumulation for atmospheric light and medium transmission will cause the deviation in color fidelity and brightness. Therefore, we designed an encoder-decoder structure for irradiance guidance, which not only eliminates error accumulation but also enhances the detail in the restored image, achieving higher-quality dehazing results. Quantitative and qualitative evaluations indicate that our dehazing method outperforms existing techniques, effectively eliminating haze from drone images and significantly enhancing image clarity and quality in hazy conditions. Specifically, the compared experiment on the R100 dataset demonstrates that the proposed method improved the peak signal-to-noise ratio (PSNR) and structure similarity index measure (SSIM) metrics by 6.9 dB and 0.08 over the second-best method, respectively. On the N100 dataset, the method improved the PSNR and SSIM metrics by 8.7 dB and 0.05 over the second-best method, respectively.

YOLO-DroneMS: Multi-Scale Object Detection Network for Unmanned Aerial Vehicle (UAV) Images

In recent years, research on Unmanned Aerial Vehicles (UAVs) has developed rapidly. Compared to traditional remote-sensing images, UAV images exhibit complex backgrounds, high resolution, and large differences in object scales. Therefore, UAV object detection is an essential yet challenging task. This paper proposes a multi-scale object detection network, namely YOLO-DroneMS (You Only Look Once for Drone Multi-Scale Object), for UAV images. Targeting the pivotal connection between the backbone and neck, the Large Separable Kernel Attention (LSKA) mechanism is adopted with the Spatial Pyramid Pooling Factor (SPPF), where weighted processing of multi-scale feature maps is performed to focus more on features. And Attentional Scale Sequence Fusion DySample (ASF-DySample) is introduced to perform attention scale sequence fusion and dynamic upsampling to conserve resources. Then, the faster cross-stage partial network bottleneck with two convolutions (named C2f) in the backbone is optimized using the Inverted Residual Mobile Block and Dilated Reparam Block (iRMB-DRB), which balances the advantages of dynamic global modeling and static local information fusion. This optimization effectively increases the model’s receptive field, enhancing its capability for downstream tasks. By replacing the original CIoU with WIoUv3, the model prioritizes anchoring boxes of superior quality, dynamically adjusting weights to enhance detection performance for small objects. Experimental findings on the VisDrone2019 dataset demonstrate that at an Intersection over Union (IoU) of 0.5, YOLO-DroneMS achieves a 3.6% increase in mAP@50 compared to the YOLOv8n model. Moreover, YOLO-DroneMS exhibits improved detection speed, increasing the number of frames per second (FPS) from 78.7 to 83.3. The enhanced model supports diverse target scales and achieves high recognition rates, making it well-suited for drone-based object detection tasks, particularly in scenarios involving multiple object clusters.

The Performance of the Optical Flow Field based Dense Image Matching for UAV Imagery

Abstract. With the rapid development of sensing platforms, unmanned aerial vehicle (UAV)-based mapping has become increasingly popular because of its economic efficiency and flexibility, especially for providing 3D information to support urban growth monitoring and change detection to meet sustainable development goals (SDGs). This paper presents an improved optical flow field-based dense matching algorithm (OFFDM) for low-altitude UAV images based on the Ph.D. thesis of Yuan (Yuan, 2018). First, high-precision seed points were used to compute the optical flow field within stereo pairs, effectively minimizing redundant calculations during the fine-matching phase. Second, a fine-matching approach, integrating multiple constraints, was applied to refine the coarse matching results based on the optical flow field. Extensive dense matching experiments on UAV low-altitude aerial imagery assessed the performance of OFFDIM across four dimensions: 3D point cloud visualization, matching success rate, precision, and reliability. Extensive experiments on low-altitude UAV imagery, characterized by a resolution of 7cm per pixel over a 10,608×8,608 pixel dimension and a 60% forward overlap, evaluate the OFFDM's efficacy. The quantitative evaluation revealed that the proposed method achieved an accuracy of ±0.7 pixels in image coordinates and ±20 cm on the ground, with a matching success rate exceeding 97%. The processing time was approximately 272 seconds for handling one single stereo pair. When compared to the widely adopted PMVS algorithm, known for its effectiveness in dense matching for UAV images, the proposed method demonstrated higher completeness and improved matching efficiency by more than five times. These results demonstrated that the proposed approach is more suitable for dense matching on UAV imagery-based high-precision 3D spatial data extraction, supporting global mapping tasks more effectively.

FG-UNet: fine-grained feature-guided UNet for segmentation of weeds and crops in UAV images.

Semantic segmentation of weed and crop images is a key component and prerequisite for automated weed management. For weeds in unmanned aerial vehicle (UAV) images, which are usually characterized by small size and easily confused with crops at early growth stages, existing semantic segmentation models have difficulties to extract sufficiently fine features. This leads to their limited performance in weed and crop segmentation of UAV images. We proposed a fine-grained feature-guided UNet, named FG-UNet, for weed and crop segmentation in UAV images. Specifically, there are two branches in FG-UNet, namely the fine-grained feature branch and the UNet branch. In the fine-grained feature branch, a fine feature-aware (FFA) module was designed to mine fine features in order to enhance the model's ability to segment small objects. In the UNet branch, we used an encoder-decoder structure to realize high-level semantic feature extraction in images. In addition, a contextual feature fusion (CFF) module was designed for the fusion of the fine features and high-level semantic features, thus enhancing the feature discrimination capability of the model. The experimental results showed that our proposed FG-UNet, achieved state-of-the-art performance compared to other semantic segmentation models, with mean intersection over union (MIOU) and mean pixel accuracy (MPA) of 88.06% and 92.37%, respectively. The proposed method in this study lays a solid foundation for accurate detection and intelligent management of weeds. It will have a positive impact on the development of smart agriculture. © 2024 Society of Chemical Industry.

Investigation, Evaluation, and Dynamic Monitoring of Traditional Chinese Village Buildings Based on Unmanned Aerial Vehicle Images and Deep Learning Methods

The investigation, evaluation, and dynamic monitoring of traditional village buildings are crucial for the protection and inheritance of their architectural styles. This study takes traditional villages in Shandong Province, China, as an example, employing UAV images and deep learning technology. Utilizing the YOLOv8 instance segmentation model, it introduces three key features reflecting the condition of traditional village buildings: roof status, roof form, and courtyard vegetation coverage. By extracting feature data on the condition of traditional village buildings and constructing a transition matrix for building condition changes, combined with corresponding manual judgment assistance, the study classifies, counts, and visually outputs the conditions and changes of buildings. This approach enables the investigation, evaluation, and dynamic monitoring of traditional village buildings. The results show that deep learning technology significantly enhances the efficiency and accuracy of traditional village architectural investigation and evaluations, and it performs well in dynamic monitoring of building condition changes. The “UAV image + deep learning” technical system, with its simplicity, accuracy, efficiency, and low cost, can provide further data and technical support for the planning, protection supervision, and development strategy formulation of traditional Chinese villages.