Multi-Vehicle Object Recognition Method Based on YOLOv7-W
To meet the demands of intelligent traffic control systems for high-precision vehicle object recognition, an improved recognition algorithm, YOLOv7-W, is proposed based on the YOLOv7 algorithm. This algorithm effectively addresses three key challenges: high vehicle miss detection rates, inadequate p...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11004061/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | To meet the demands of intelligent traffic control systems for high-precision vehicle object recognition, an improved recognition algorithm, YOLOv7-W, is proposed based on the YOLOv7 algorithm. This algorithm effectively addresses three key challenges: high vehicle miss detection rates, inadequate perception of small-angle targets, and insufficient feature extraction capabilities. To enhance YOLOv7’s feature extraction abilities, the Global Attention Mechanism (GAM) is integrated into the feature extraction process. To improve the perception of small targets, the backbone network is reconstructed using Omni-dimensional Dynamic and Efficient Aggregation network module (ODEANet) to capture more comprehensive global information. The Context-aware Transformer (CoT) block is adopted to replace the original E-ELAN module in the neck network, guiding dynamic attention matrix learning and optimizing visual representations. Meanwhile, the computational complexity in terms of GFLOPs (Giga Floating Point Operations) is reduced by 3.1%. Analysis of the XUPEI-CAR experimental dataset reveals significant variations in features across different traffic flow densities. To improve the matching accuracy of prior frames, the k-means++ clustering algorithm is employed to optimize the prior frame parameters. Experimental results demonstrate that YOLOv7-W achieves recognition accuracies of 97.125%, 94.85%, and 94.6% in free-flow, synchronized flow, and congested traffic scenarios, respectively. Compared to YOLOv7, these improvements result in accuracy increases of 3.65%, 3.2%, and 1.4% in the three scenarios. The optimized architecture maintains real-time performance while reducing model parameters by 13.37%, with frame rates of 74.63, 79.37, and 75.76 FPS in different traffic conditions. |
|---|---|
| ISSN: | 2169-3536 |