Heuristic Optimal Scheduling for Road Traffic Incident Detection Under Computational Constraints

Area of Science:

• Computer Science
• Artificial Intelligence
• Traffic Engineering

Background:

• Intelligent monitoring of road surveillance videos is vital for traffic anomaly detection and road safety.
• Current methods face challenges with high computational demands or overlooking events due to periodic analysis.
• Existing static resource allocation models often lead to wastage or insufficiency.

Purpose of the Study:

• To introduce a heuristic optimal scheduling approach for efficient traffic event detection under limited computational resources.
• To enhance the effectiveness of traffic anomaly detection and early warning systems.
• To address the limitations of traditional real-time analysis and periodic camera patrol methods.

Main Methods:

• A heuristic optimal scheduling approach is proposed, leveraging historical data and prior knowledge.
• A weighted event feature value is computed for each camera to quantify detection efficiency.
• A cyclic elimination mechanism dynamically reallocates resources from low-performing to high-performing cameras.

Main Results:

• The proposed method demonstrated substantial improvements in traffic event detection efficiency (40%, 28%, 17%, 28% across periods).
• Outperformed existing resource scheduling algorithms in average load degree, load balance degree, and computational resource utilization.
• Validated through a case study in a major metropolitan city in China.

Conclusions:

• The heuristic optimal scheduling approach effectively enhances traffic event detection efficiency within limited computational resources.
• This method offers greater flexibility and adaptability compared to static allocation models.
• The approach successfully addresses practical demands for traffic anomaly detection and early warning systems.