Recognition new energy vehicles based on improved YOLOv5

Area of Science:

• Intelligent Traffic Systems (ITS)
• Computer Vision
• Machine Learning

Background:

• Vehicle recognition is crucial for ITS, but distinguishing new energy vehicles in complex environments remains challenging.
• Existing methods struggle with accurate identification and traffic flow statistics for diverse vehicle types.

Purpose of the Study:

• To propose and validate a novel approach for New Energy Vehicle Recognition and Traffic Flow Statistics (NEVTS).
• To enhance vehicle detection and classification accuracy in intelligent traffic systems.

Main Methods:

• Utilized You Only Look Once v5 (YOLOv5) for vehicle detection and license plate recognition.
• Implemented Task-Specific Context Decoupling (TSCODE) to improve YOLOv5's prediction and classification performance.
• Classified vehicles based on license plate color (green for new energy, non-green for fuel).

Main Results:

• The NEVTS approach significantly improved vehicle detection performance.
• Accurate and efficient identification of new energy vehicles and traffic flow statistics was achieved.
• Experimental results demonstrated the effectiveness of the proposed method.

Conclusions:

• The NEVTS approach provides an effective solution for new energy vehicle recognition and traffic flow statistics.
• This method can be extended for traffic pattern analysis and management in intelligent traffic systems.