Multiple object tracking using the shortest path faster association algorithm.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Operations Research

Background:

• Multiple object tracking (MOT) in cluttered environments remains a significant challenge.
• Existing methods often struggle with robustness and computational efficiency.

Purpose of the Study:

• To develop a novel and efficient tracking association approach for multiple object tracking.
• To improve tracking accuracy and robustness in complex, cluttered scenarios.

Main Methods:

• Formulating multiple object tracking as an integer programming problem within a flow network.
• Relaxing the integer programming to a standard linear programming problem.
• Employing the shortest path faster algorithm to efficiently find the global optimum.

Main Results:

• The proposed method demonstrates lower worst-case complexity compared to competing algorithms.
• Achieved superior robustness and tracking accuracy in complex environments.
• Simulation results confirm reduced computation time, enabling real-time operation.

Conclusions:

• The shortest path faster algorithm offers an effective solution for multiple object tracking.
• This approach overcomes limitations of traditional integer programming methods.
• The algorithm provides a computationally efficient and accurate solution for real-time tracking applications.