Optimal fuzzy-PID controller design for object tracking

Area of Science:

• Computer Vision
• Robotics
• Control Systems

Background:

• Object tracking is crucial for applications like video surveillance and traffic monitoring.
• Existing methods face challenges with dynamic camera movements and varying object speeds.

Purpose of the Study:

• To develop and compare static and dynamic camera-based object tracking techniques.
• To design an optimal Fuzzy-PID controller for dynamic camera object tracking.
• To evaluate system performance across various object trajectories and frequencies.

Main Methods:

• Static camera tracking implemented using NI LabVIEW with a Shape Adaptive Mean-Shift algorithm.
• Dynamic camera tracking utilized an optimal Fuzzy-PID controller tuned by a Genetic Algorithm (GA).
• System performance evaluated with step, sinusoidal, circular, and elliptical trajectories at 1, 50, and 100 rad/sec.

Main Results:

• The GA-tuned Fuzzy-PID controller significantly reduced steady-state error and improved rise/settling times compared to PID, Fuzzy, and standard Fuzzy-PID.
• Optimal controller demonstrated superior object position stabilization.
• System performance was best at lower object frequencies, decreasing as speed increased.

Conclusions:

• The developed optimal Fuzzy-PID controller is highly effective and appropriate for dynamic camera object tracking.
• The system's performance degradation at higher speeds aligns with real-world tracking limitations.
• This research contributes to more robust and efficient object tracking solutions.