Robust Vehicle Detection and Counting Algorithm Employing a Convolution Neural Network and Optical Flow.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Traffic Engineering

Background:

• Effective traffic control and management rely on accurate vehicle detection and counting.
• Existing methods struggle with complex traffic scenes, necessitating improved algorithms.

Purpose of the Study:

• To develop an effective algorithm for vehicle detection and counting in complex traffic scenes.
• To enhance traffic management systems through improved automated vehicle analysis.

Main Methods:

• Combines convolution neural network (CNN) for detection with optical flow feature tracking (KLT tracker and K-means clustering) for refinement.
• Integrates detection and tracking for robust feature points updated regularly.
• Utilizes background subtraction with CNN for initial vehicle identification.

Main Results:

• Achieved an average detection precision of 96.3% and counting precision of 96.8%.
• Demonstrated superior performance compared to existing approaches in challenging video environments.
• Successfully assigned vehicle labels to trajectories and accurately counted vehicles.

Conclusions:

• The proposed hybrid CNN and optical flow algorithm offers a robust solution for vehicle detection and counting.
• This method significantly enhances accuracy in complex traffic scenarios.
• The findings contribute to more efficient traffic control and management systems.