Color Feature-Based Object Tracking through Particle Swarm Optimization with Improved Inertia Weight.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Object tracking is crucial in computer vision.
• Traditional particle swarm optimization (PSO) algorithms face limitations in object tracking, particularly with occlusion and non-rigid deformations.
• Existing inertia weight adjustment mechanisms in PSO lack adaptability to particle states.

Purpose of the Study:

• To develop an improved particle swarm tracking algorithm robust to deformations, scale variations, rotations, and partial occlusions.
• To enhance the inertia weight adjustment mechanism in PSO for adaptive tracking.
• To achieve state-of-the-art performance in object tracking scenarios.

Main Methods:

• Utilizing a weighted color histogram as the target feature to mitigate the impact of edge pixels.
• Implementing an improved inertia weight adjustment mechanism based on the concept of 'particle maturity'.
• Applying particle swarm optimization (PSO) for multi-peak search to handle target occlusion.

Main Results:

• The proposed algorithm demonstrates insensitivity to non-rigid deformation, scale variation, and rotation.
• Reduced influence of partial obstruction on target feature description.
• Experimental results indicate state-of-the-art performance across diverse tracking scenarios.

Conclusions:

• The enhanced particle swarm tracking algorithm provides accurate and robust object tracking.
• The 'particle maturity' concept effectively improves inertia weight adaptation in PSO for tracking.
• The algorithm shows significant improvements in handling challenging tracking conditions.