A New Multi-Branch Convolutional Neural Network and Feature Map Extraction Method for Traffic Congestion Detection.

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Engineering

Background:

• Increasing vehicle numbers and economic advancement exacerbate traffic congestion on key roadways.
• Existing traffic management systems often lack efficient, automated congestion detection methods.
• Leveraging existing camera networks for traffic analysis presents a cost-effective solution.

Purpose of the Study:

• To develop a deep learning model for automatic traffic congestion detection using camera image data.
• To introduce a new Vehicle Information Feature Map (VIFM) method and a Multi-Branch Convolutional Neural Network (MBCNN) model.
• To provide an effective deep learning-based solution for traffic congestion detection without significant hardware investment.

Main Methods:

• Vehicle detection using an object detection model.
• Extraction of a novel Vehicle Information Feature Map (VIFM).
• Development of a traffic congestion detection model utilizing a Multi-Branch Convolutional Neural Network (MBCNN).

Main Results:

• The proposed VIFM and MBCNN method demonstrated superior performance compared to existing models.
• Achieved a high F1 score of 98.61% and an accuracy of 98.62% on the CCTRIB dataset.
• Validated the effectiveness of the deep learning approach in real-world traffic image data.

Conclusions:

• The developed deep learning model effectively detects traffic congestion from camera images.
• The VIFM and MBCNN method offers a powerful and efficient tool for traffic management.
• This approach enables the utilization of existing transportation camera infrastructure for enhanced traffic monitoring.