Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 227, P. 109646 - 109646
Published: Nov. 25, 2024
Language: Английский
PLoS ONE, Journal Year: 2025, Volume and Issue: 20(1), P. e0312112 - e0312112
Published: Jan. 8, 2025
The quality and safety of tea food production is paramount importance. In traditional processing techniques, there a risk small foreign objects being mixed into Pu-erh sun-dried green tea, which directly affects the food. To rapidly detect accurately identify these in this study proposes an improved YOLOv8 network model for object detection. method employs MPDIoU optimized loss function to enhance target detection performance, thereby increasing model’s precision targeting. It incorporates EfficientDet high-efficiency architecture module, utilizes compound scale-centered anchor boxes adaptive feature pyramid achieve efficient targets various sizes. BiFormer bidirectional attention mechanism introduced, allowing consider both forward backward dependencies sequence data, significantly enhancing understanding context images. further integrated with sliced auxiliary super-inference technology YOLOv8, subdivides image conducts in-depth analysis local features, improving recognition accuracy robustness multi-scale objects. Experimental results demonstrate that, compared original model, has seen increases 4.50% Precision, 5.30% Recall, 3.63% mAP, 4.9% F1 score. When YOLOv7, YOLOv5, Faster-RCNN, SSD models, its by 3.92%, 7.26%, 14.03%, 11.30%, respectively. This research provides new technological means intelligent transformation automated color sorters, equipment, sorting systems high-quality Yunnan tea. also strong technical support automation development industry.
Language: Английский
Citations
1
Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)
Published: Feb. 8, 2025
Language: Английский
Citations
0
Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 234, P. 110221 - 110221
Published: March 13, 2025
Language: Английский
Citations
0
Lecture notes in computer science, Journal Year: 2025, Volume and Issue: unknown, P. 291 - 305
Published: Jan. 1, 2025
Language: Английский
Citations
0
Agriculture, Journal Year: 2024, Volume and Issue: 14(7), P. 1011 - 1011
Published: June 26, 2024
Accurately identifying cherry-tomato picking points and obtaining their coordinate locations is critical to the success of robots. However, previous methods for semantic segmentation alone or combining object detection with traditional image processing have struggled accurately determine point due challenges such as leaves well targets that are too small. In this study, we propose a YOLOv8n-DDA-SAM model adds branch target achieve desired compute point. To be specific, YOLOv8n used initial model, dynamic snake convolutional layer (DySnakeConv) more suitable stems in neck model. addition, large kernel attention mechanism adopted backbone use ADown convolution resulted better fusion stem features certain decrease number parameters without loss accuracy. Combined SAM, mask effectively obtained then accurate by simple shape-centering calculation. As suggested experimental results, proposed significantly improved from models not only detecting but also stem’s masks. [email protected] F1-score, achieved 85.90% 86.13% respectively. Compared original YOLOv8n, YOLOv7, RT-DETR-l YOLOv9c, has 24.7%, 21.85%, 19.76%, 15.99% F1-score increased 16.34%, 12.11%, 10.09%, 8.07% respectively, 6.37M. branch, does it need produce relevant datasets, its mIOU 11.43%, 6.94%, 5.53%, 4.22% 12.33%, 7.49%, 6.4%, 5.99% compared Deeplabv3+, Mask2former, DDRNet SAN summary, can satisfy requirements high-precision provides strategy system cherry-tomato.
Language: Английский
Citations
3
Sensors, Journal Year: 2024, Volume and Issue: 24(17), P. 5586 - 5586
Published: Aug. 28, 2024
Road crack detection is of paramount importance for ensuring vehicular traffic safety, and implementing traditional methods cracks inevitably impedes the optimal functioning traffic. In light above, we propose a USSC-YOLO-based target algorithm unmanned aerial vehicle (UAV) road based on machine vision. The aims to achieve high-precision at all scale levels. Compared with original YOLOv5s, main improvements USSC-YOLO are ShuffleNet V2 block, coordinate attention (CA) mechanism, Swin Transformer. First, address problem large network computational spending, replace backbone YOLOv5s blocks, reducing overhead significantly. Next, reduce problems caused by complex background interference, introduce CA mechanism into network, which reduces missed false rate. Finally, integrate Transformer block end neck enhance accuracy small cracks. Experimental results our self-constructed UAV near-far scene i(UNFSRCI) dataset demonstrate that model giga floating-point operations per second (GFLOPs) compared while achieving 6.3% increase in mAP@50 12% improvement mAP@ [50:95]. This indicates remains lightweight meanwhile providing excellent performance. future work, will assess safety conditions these prioritize maintenance sequences targets facilitate further intelligent management.
Language: Английский
Citations
3
Sensors, Journal Year: 2024, Volume and Issue: 24(21), P. 6982 - 6982
Published: Oct. 30, 2024
Detecting faulty pipelines in water management systems is crucial for ensuring a reliable supply of clean water. Traditional inspection methods are often time-consuming, costly, and prone to errors. This study introduces an AI-based model utilizing images detect pipeline defects, focusing on leaks, cracks, corrosion. The YOLOv8 employed object detection due its exceptional performance detecting objects, segmentation, pose estimation, tracking, classification. By training large dataset labeled images, the effectively learns identify visual patterns associated with faults. Experiments conducted real-world demonstrate that significantly outperforms traditional accuracy. also exhibits robustness various environmental conditions such as lighting changes, camera angles, occlusions, diverse scenarios. efficient processing time enables real-time fault large-scale distribution networks implementing this offers numerous advantages systems. It reduces dependence manual inspections, thereby saving costs enhancing operational efficiency. Additionally, facilitates proactive maintenance through early faults, preventing loss, contamination, infrastructure damage. results from three experiments indicate Experiment 1 achieves commendable mAP50 90% pipes, overall 74.7%. In contrast, 3 superior performance, achieving 76.1%. research presents promising approach improving reliability sustainability using image analysis.
Language: Английский
Citations
3
Agriculture, Journal Year: 2024, Volume and Issue: 14(8), P. 1346 - 1346
Published: Aug. 12, 2024
Berries are nutritious and valuable, but their thin skin, soft flesh, fragility make harvesting picking challenging. Manual traditional mechanical methods commonly used, they costly in labor can damage the fruit. To overcome these challenges, it may be worth exploring alternative methods. Using berry fruit-picking robots with perception technology is a viable option to improve efficiency of harvesting. This review presents an overview mechanisms robots, encompassing underlying principles, mechanics grasping, examination structural design. The importance during process highlighted. Then, several techniques used by described, including visual perception, tactile distance measurement, switching sensors. four perceptual berry-picking advantages disadvantages analyzed. In addition, technical characteristics technologies practical applications analyzed summarized, advanced presented. Finally, challenges that need prospects for overcoming discussed.
Language: Английский
Citations
2
Sensors, Journal Year: 2024, Volume and Issue: 24(17), P. 5789 - 5789
Published: Sept. 5, 2024
Many compartments are prone to pose safety hazards such as loose fasteners or object intrusion due their confined space, making manual inspection challenging. To address the challenges of complex environments, diverse target categories, and variable scales in compartments, this paper proposes a novel GMS-YOLO network, based on improved YOLOv8 framework. In addition lightweight design, network accurately detects targets by leveraging more precise high-level low-level feature representations obtained from GhostHGNetv2, which enhances feature-extraction capabilities. handle issue backbone employs GhostHGNetv2 capture accurate representations, facilitating better distinction between background targets. addition, significantly reduces both parameter size computational complexity. varying scales, first layer fusion module introduces Multi-Scale Convolutional Attention (MSCA) multi-scale contextual information guide process. A new detection head, Shared Detection Head (SCDH), is designed enable model achieve higher accuracy while being lighter. evaluate performance algorithm, dataset for scenario was constructed. The experiment results indicate that compared original model, number decreased 37.8%, GFLOPs 27.7%, average increased 82.7% 85.0%. This validates applicability proposed network.
Language: Английский
Citations
2
Journal of Marine Science and Engineering, Journal Year: 2024, Volume and Issue: 12(9), P. 1641 - 1641
Published: Sept. 13, 2024
The detection and surveillance of ship targets in coastal waters is not only a crucial technology for the advancement intelligence, but also holds great significance safety economic development areas. However, due to poor visibility foggy conditions, effectiveness during weather limited. In this paper, we propose an improved version YOLOv8s, termed YOLOv8s-Fog, which provides multi-target network specifically designed nearshore scenes weather. This improvement involves adding coordinate attention neck YOLOv8 replacing convolution C2f with deformable convolution. Additionally, expand dataset, construct synthesize collection target images captured on days varying degrees fog, using atmospheric scattering model monocular depth estimation. We compare standard YOLOv8s model, as well several other object models. results demonstrate superior performance achieved by achieving average accuracy 74.4% ([email protected]), 1.2% higher than that model.
Language: Английский
Citations
2