G-YOLO: A YOLOv7-based target detection algorithm for lightweight hazardous chemical vehicles.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Transportation Safety

Background:

• Hazardous chemical vehicles transport dangerous substances, posing risks of fire, explosion, and leakage.
• Ensuring safety during the transportation of hazardous materials is critical for human and environmental protection.
• Existing object detection methods may lack efficiency for real-time applications in this domain.

Purpose of the Study:

• To develop a lightweight and efficient object detection method for hazardous chemical vehicles.
• To improve the accuracy and robustness of detecting hazardous chemical vehicles.
• To reduce the computational burden and parameter count of object detection models.

Main Methods:

• Utilized a lightweight feature extraction structure (E-GhostV2 network) in the YOLOv7-tiny model's trunk and neck.
• Incorporated Partial Convolution (PConv) in the model's backbone to reduce computations and memory access.
• Employed the WIoU loss function to balance training on high-quality and low-quality samples, enhancing generalization.

Main Results:

• The proposed method achieves satisfactory detection accuracy for hazardous chemical vehicles.
• The improved model significantly reduces the number of model parameters compared to the baseline.
• Enhanced efficiency and feature extraction capabilities were observed.

Conclusions:

• The lightweight object detection model provides a robust solution for identifying hazardous chemical vehicles.
• The method offers practical support for enhancing safety and theoretical research in hazardous material transport.
• The optimized model balances detection performance with computational efficiency.