YOLO-UAV: Object Detection Method of Unmanned Aerial Vehicle Imagery Based on Efficient Multi-Scale Feature Fusion

Ma, Chengji; Fu, Yanyun; Wang, Deyong; Guo, Rui; Zhao, Xueyi; Fang, Jian

doi:10.1109/access.2023.3329713

Cited by 8 publications

(3 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main contributions of this paper are as follows: (5) With the improved YOLOv8 model, we achieved an accuracy of 97.59%, a recall rate of 94.40%, a frame rate of 40.62 FPS, model parameters of 14.62 MB, and GFLOPs of 11.21. These results demonstrate the model's capability to undertake large-scale corn tassel recognition and counting tasks with UAVs under complex conditions [22].…”

Section: Contribution Of This Articlementioning

confidence: 71%

“…Drones 2024, 8, x FOR PEER REVIEW 5 of 24 (4) We proposed a learning rate optimization method based on the sparrow search algorithm (SSA) to obtain the optimal learning rate for the highest average precision, thereby enhancing the model's robustness and detection accuracy; (5) With the improved YOLOv8 model, we achieved an accuracy of 97.59%, a recall rate of 94.40%, a frame rate of 40.62 FPS, model parameters of 14.62 MB, and GFLOPs of 11.21. These results demonstrate the model's capability to undertake large-scale corn tassel recognition and counting tasks with UAVs under complex conditions [22].…”

Section: Acquisition Of Corn Tassel Imagesmentioning

confidence: 71%

See 1 more Smart Citation

Multi-Altitude Corn Tassel Detection and Counting Based on UAV RGB Imagery and Deep Learning

Niu,

Nie,

et al. 2024

Drones

View full text Add to dashboard Cite

In the context of rapidly advancing agricultural technology, precise and efficient methods for crop detection and counting play a crucial role in enhancing productivity and efficiency in crop management. Monitoring corn tassels is key to assessing plant characteristics, tracking plant health, predicting yield, and addressing issues such as pests, diseases, and nutrient deficiencies promptly. This ultimately ensures robust and high-yielding corn growth. This study introduces a method for the recognition and counting of corn tassels, using RGB imagery captured by unmanned aerial vehicles (UAVs) and the YOLOv8 model. The model incorporates the Pconv local convolution module, enabling a lightweight design and rapid detection speed. The ACmix module is added to the backbone section to improve feature extraction capabilities for corn tassels. Moreover, the CTAM module is integrated into the neck section to enhance semantic information exchange between channels, allowing for precise and efficient positioning of corn tassels. To optimize the learning rate strategy, the sparrow search algorithm (SSA) is utilized. Significant improvements in recognition accuracy, detection efficiency, and robustness are observed across various UAV flight altitudes. Experimental results show that, compared to the original YOLOv8 model, the proposed model exhibits an increase in accuracy of 3.27 percentage points to 97.59% and an increase in recall of 2.85 percentage points to 94.40% at a height of 5 m. Furthermore, the model optimizes frames per second (FPS), parameters (params), and GFLOPs (giga floating point operations per second) by 7.12%, 11.5%, and 8.94%, respectively, achieving values of 40.62 FPS, 14.62 MB, and 11.21 GFLOPs. At heights of 10, 15, and 20 m, the model maintains stable accuracies of 90.36%, 88.34%, and 84.32%, respectively. This study offers technical support for the automated detection of corn tassels, advancing the intelligence and precision of agricultural production and significantly contributing to the development of modern agricultural technology.

show abstract

Section: Contribution Of This Articlementioning

confidence: 71%

Section: Acquisition Of Corn Tassel Imagesmentioning

confidence: 71%

Multi-Altitude Corn Tassel Detection and Counting Based on UAV RGB Imagery and Deep Learning

Niu,

Nie,

et al. 2024

Drones

View full text Add to dashboard Cite

show abstract

“…Ma [32] et al proposed the GS decoupling head, which separates classification tasks from regression tasks to reduce the impact of task differences on prediction bias. Additionally, a new loss function Focal-ECIoU was introduced to accelerate network convergence and enhance the model's localization capability.…”

Section: Introductionmentioning

confidence: 99%

L-ENet: An Ultralightweight SAR Image Detection Network

Wang,

Miao,

Zhu

2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Synthetic Aperture Radar (SAR) imagery plays a vital role in maritime vessel detection, facing challenges like complex marine environments, low-contrast target recognition, and similarities between ships and sea surface waves. Additionally, the deployment of traditional large models on radar-equipped edge devices is hampered by their extensive computational demands. Addressing these challenges, this study presents an ultra-lightweight network Lightweight Enhanced Network(L-ENet), improving upon YOLOv5n. The backbone network is switched to a more efficient ShufflenetV2, integrated with an improved attention mechanism, DCFAM (Dual-Core Fusion Attention Mechanism), for effective inter-channel information balancing. In the network's neck, the original C3 module has been replaced with the C3EGhost convolution. This convolution module integrates the Ghost and ECA (Efficient Channel Attention) attention mechanisms, aiming to mitigate potential accuracy loss during the lightweight process. Furthermore, this study introduces a novel multi-scale fusion pathway and a Concat module with adaptive weights to better harmonize semantic and detail information. Further, the tri-head detection structure is revised to a dual-head structure OMNI-DIMENSIONAL Adaptive Spatial Feature Fusion (ODASFF), using ODConv to allocate weights across different scales for efficient detection. Experimental results show that L-ENet has a computational cost of 0.6M and a parameter count of 2.1GFLOPs. These figures are 65% and 50% lower than those of the original YOLOv5n while maintaining the same detection accuracy of 97.8%. Hence, L-ENet offers an efficient and viable solution for maritime target detection, showcasing commendable detection performance in comparison with advanced models.

show abstract

DMFF-YOLO: YOLOv8 Based on Dynamic Multiscale Feature Fusion for Object Detection on UAV Aerial Photography

Qiu,

Chen,

Sun

et al. 2024

IEEE Access

View full text Add to dashboard Cite

YOLO-UAV: Object Detection Method of Unmanned Aerial Vehicle Imagery Based on Efficient Multi-Scale Feature Fusion

Cited by 8 publications

References 52 publications

Multi-Altitude Corn Tassel Detection and Counting Based on UAV RGB Imagery and Deep Learning

Multi-Altitude Corn Tassel Detection and Counting Based on UAV RGB Imagery and Deep Learning

L-ENet: An Ultralightweight SAR Image Detection Network

DMFF-YOLO: YOLOv8 Based on Dynamic Multiscale Feature Fusion for Object Detection on UAV Aerial Photography

Contact Info

Product

Resources

About