Comprehensive Investigation of Unmanned Aerial Vehicles (UAVs): An In-Depth Analysis of Avionics Systems

Sensors, Journal Year: 2024, Volume and Issue: 24(10), P. 3064 - 3064

Published: May 11, 2024

The evolving technologies regarding Unmanned Aerial Vehicles (UAVs) have led to their extended applicability in diverse domains, including surveillance, commerce, military, and smart electric grid monitoring. Modern UAV avionics enable precise aircraft operations through autonomous navigation, obstacle identification, collision prevention. structures of are generally complex, thorough hierarchies intricate connections exist between. For a comprehensive understanding design, this paper aims assess critically review the purpose-classified electronics hardware inside UAVs, each with corresponding performance metrics thoroughly analyzed. This includes an exploration different algorithms used for data processing, flight control, protection, communication. Consequently, enriches knowledge base offering informative background on various design processes, particularly those related applications. As future work recommendation, actual relevant project is openly discussed.

Language: Английский

---

Application of Convolutional Neural Networks in Weed Detection and Identification: A Systematic Review.

Published: Jan. 16, 2024

Weeds are unwanted and invasive plants that proliferate compete for resources such as space, water, nutrients, sunlight, affecting the quality productivity of desired crops. Weed detection is crucial application precision agriculture methods this purpose machine learning techniques can be used, specifically convolutional neural networks (CNN). This study focuses on search CNN architectures technology used to detect identify weeds in different crops; 61 articles applying were analyzed last five years (2019-2023). The results show devices acquire image training, digital cameras, smartphones, drone cameras. Additionally, YOLO family algorithms most widely adopted architectures, followed by VGG, ResNet, Faster R-CNN, AlexNet, MobileNet, respectively. provides an update CNNs will serve a starting point researchers wishing implement these weed identification techniques.

Language: Английский

---

Application of Convolutional Neural Networks in Weed Detection and Identification: A Systematic Review

Agriculture, Journal Year: 2024, Volume and Issue: 14(4), P. 568 - 568

Published: April 2, 2024

Weeds are unwanted and invasive plants that proliferate compete for resources such as space, water, nutrients, sunlight, affecting the quality productivity of desired crops. Weed detection is crucial application precision agriculture methods this purpose machine learning techniques can be used, specifically convolutional neural networks (CNN). This study focuses on search CNN architectures used to detect identify weeds in different crops; 61 articles applying were analyzed during last five years (2019–2023). The results show devices acquire images training, digital cameras, smartphones, drone cameras. Additionally, YOLO family algorithms most widely adopted architectures, followed by VGG, ResNet, Faster R-CNN, AlexNet, MobileNet, respectively. provides an update CNNs will serve a starting point researchers wishing implement these weed identification techniques.

Language: Английский

---

Weed species classification with UAV imagery and standard CNN models: Assessing the frontiers of training and inference phases

Crop Protection, Journal Year: 2024, Volume and Issue: 182, P. 106721 - 106721

Published: May 11, 2024

Accurate weed species identification is crucial for effective site-specific management (SSWM), enabling targeted and timely control measures each in crop field. This study advanced the current approach to species-level during early growth stage by integrating unmanned aerial vehicles (UAVs) imagery with standard convolutional neural networks (CNNs) models such as VGG16, Resnet152 Inception-Resnet-v2. For this, a robust dataset was created 33,467 labels of weeds (Atriplex patula, Chenopodium album, Convolvulus arvensis, Cyperus rotundus, Lolium rigidum, Portulaca oleracea, Salsola kali, Solanum nigrum) crops (maize, tomato), which subjected different training, validation test scenarios. Model inputs were adjusted order align them information represented UAV images. Initially, developed balanced scenarios, gradually increasing label numbers assess their performance. Inception-ResNet-v2 achieved over 90% accuracy 400 labels, while ResNet152 VGG16 required 600 800 respectively, similar accuracy. In more complex realistic scenarios unbalanced datasets, outperformed, likely due its deeper architecture enhanced capability capture intricate features patterns within The emphasized importance minority-to-majority ratio affects minority classification. To prevent misclassification, it determine right number CNN model training validation. Weed maps generated after classification using Faster R-CNN algorithm an object detector. advancement methodology facilitates precise efficient implementation SSWM techniques.

Language: Английский

---

Overhead power line detection from aerial images using segmentation approaches

Automatika, Journal Year: 2024, Volume and Issue: 65(1), P. 261 - 288

Published: Jan. 2, 2024

Ensuring the optimal efficiency of electrical networks requires vigilant surveillance and preventive maintenance. While traditional methods, such as human patrols helicopter inspections, have been longstanding practices for grid control by power distribution companies, emergence Unmanned Aerial Vehicles (UAV) technology offers a more efficient technologically advanced alternative. The proposed comprehensive pipeline integrates various elements, including preprocessing techniques, deep learning (DL) models, classification algorithms (CA), Hough transform, to effectively detect powerlines in intricate aerial images characterized complex backgrounds. begins with Canny edge detection, progresses through morphological reconstruction using Otsu thresholding, concludes development RsurgeNet model. This versatile model performs binary feature extraction line identification. transform is employed extract semantic from Comparative assessments against three existing architectures highlight superior performance RsurgeNet. Experimental results on VL-IR dataset, encompassing both visible light (VL) infrared (IR) validate effectiveness approach. demonstrates reduced computational requirements, achieving heightened accuracy precision. contribution significantly enhances field network maintenance surveillance, providing an precise solution detection.

Language: Английский

---

Cognitive Computing Advancements: Improving Precision Crop Protection through UAV Imagery for Targeted Weed Monitoring

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(16), P. 3026 - 3026

Published: Aug. 18, 2024

Early detection of weeds is crucial to manage effectively, support decision-making and prevent potential crop losses. This research presents an innovative approach develop a specialized cognitive system for classifying detecting early-stage at the species level. The primary objective was create automated multiclass discrimination using computing, regardless weed growth stage. Initially, model trained tested on dataset 31,002 UAV images, including ten manually identified by experts early phenological stages maize (BBCH14) tomato (BBCH501). images were captured 11 m above ground resulted in classification accuracy exceeding 99.1% vision transformer Swin-T model. Subsequently, generative modeling employed data augmentation, resulting new models based architecture. These evaluated unbalanced 36,556 later (maize BBCH17 BBCH509), achieving weighted average F1-score ranging from 94.8% 95.3%. performance highlights system’s adaptability morphological variations its robustness diverse scenarios, suggesting that can be effectively implemented real agricultural significantly reducing time resources required identification. proposed augmentation technique also proved effective implementing architecture, improving generalization capability enabling accurate different stages. represents significant advancement monitoring across stages, with applications precision agriculture sustainable management. Furthermore, methodology showcases versatility latest generation application other knowledge domains, facilitating time-efficient development. Future could investigate applicability geographical regions types crops, as well real-time implementation continuous field monitoring.

Language: Английский

---

ADVANCING PRECISION AGRICULTURE WITH UAV’S: INNOVATIONS IN FERTILIZATION

INMATEH Agricultural Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 1057 - 1072

Published: Feb. 12, 2025

Unmanned Aerial Vehicles (UAVs) are revolutionizing precision agriculture, particularly in the domain of fertilization. Equipped with advanced sensors, mapping tools, and variable-rate application systems, drones enable farmers to precisely distribute fertilizers based on field variability. This targeted approach reduces waste, minimizes environmental impact, optimizes crop yield. The integration technologies such as multispectral imaging AI-driven decision-making systems further enhances efficiency by allowing real-time assessment soil conditions. Despite their numerous advantages, challenges high costs, regulatory limitations, technical scalability remain key barriers widespread adoption. article explores innovations UAVs bring fertilization, benefits, obstacles hindering broader agriculture

Language: Английский

---

Evaluation of Weed Infestations in Row Crops Using Aerial RGB Imaging and Deep Learning

Agriculture, Journal Year: 2025, Volume and Issue: 15(4), P. 418 - 418

Published: Feb. 16, 2025

One of the important factors negatively affecting yield row crops is weed infestations. Using non-contact detection methods allows for a rapid assessment infestations’ extent and management decisions practical control. This study aims to develop demonstrate methodology early evaluation infestations in maize using UAV-based RGB imaging pixel-based deep learning classification. An experimental was conducted determine on two tillage technologies, plowing subsoiling, tailored specific soil climatic conditions Southern Dobrudja. Based an with DeepLabV3 classification algorithm, it found that ResNet-34-backed model ensures highest performance compared different versions ResNet, DenseNet, VGG backbones. The achieved reached precision, recall, F1 score, Kappa, respectively, 0.986, 0.957. After applying field investigated higher level infestation observed subsoil deepening areas, where 4.6% area infested, 0.97% treatment. work contributes novel insights into during critical growth stages maize, providing robust framework optimizing control strategies this region.

Language: Английский

---

Simultaneous Learning Knowledge Distillation for Image Restoration: Efficient Model Compression for Drones

Drones, Journal Year: 2025, Volume and Issue: 9(3), P. 209 - 209

Published: March 14, 2025

Deploying high-performance image restoration models on drones is critical for applications like autonomous navigation, surveillance, and environmental monitoring. However, the computational memory limitations of pose significant challenges to utilizing complex in real-world scenarios. To address this issue, we propose Simultaneous Learning Knowledge Distillation (SLKD) framework, specifically designed compress resource-constrained drones. SLKD introduces a dual-teacher, single-student architecture that integrates two complementary learning strategies: Degradation Removal (DRL) Image Reconstruction (IRL). In DRL, student encoder learns eliminate degradation factors by mimicking Teacher A, which processes degraded images BRISQUE-based extractor capture degradation-sensitive natural scene statistics. Concurrently, IRL, decoder reconstructs clean from B, images, guided PIQE-based emphasizes preservation edge texture features essential high-quality reconstruction. This dual-teacher approach enables model learn both simultaneously, achieving robust while significantly reducing complexity. Experimental evaluations across five benchmark datasets three tasks—deraining, deblurring, dehazing—demonstrate that, compared teacher models, achieve an average reduction 85.4% FLOPs 85.8% parameters, with only slight decrease 2.6% PSNR 0.9% SSIM. These results highlight practicality integrating SLKD-compressed into systems, offering efficient real-time aerial platforms operating challenging environments.

Language: Английский

---

Automatic weed quantification in potato crops based on a modified convolutional neural network using drone images

Data & Metadata, Journal Year: 2025, Volume and Issue: 4, P. 194 - 194

Published: Feb. 13, 2025

Identifying and quantifying weeds is a crucial aspect of agriculture for efficiently controlling them. Weeds compete with the crop nutrients, minerals, physical space, sunlight, water, causing problems in crops ranging from low production to economic losses environmental deterioration land. Weed quantification generally manual process requiring significant time precision. Convolutional Neural Networks (CNN) are very common weed quantification. Thus, purpose this research adaptation ResNeXt50 CNN architecture semantic segmentation tasks, focused on automatic (Broadleaf dock, Dandelion, Kikuyo grass, other unidentified classes) potato fields using RGB images acquired by DJI Mavic 2 Pro drone. The analytical model was trained following Knowledge Discovery Databases (KDD) methodology Python TensorFlow-Keras frameworks. results indicate that modified presented mean IoU 0.7350, performance comparable values reported authors considering fewer classes. Student´s t-test Pearson correlation coefficient were applied contrast coverage predictions ground truth, indicating no statistically differences between both measurements most

Language: Английский

---