Image Moment-Based Extended Object Tracking for Complex Motions.

Area of Science:

• Computer Vision
• Robotics
• Signal Processing

Background:

• Extended object tracking is crucial for autonomous systems.
• Estimating object shape and trajectory from noisy data presents significant challenges.
• Existing methods often struggle with complex target motions and limited measurement points.

Purpose of the Study:

• To develop a novel image moment-based model for shape estimation and tracking of extended targets.
• To improve the accuracy and robustness of extended object tracking algorithms.
• To handle targets moving along complex trajectories using noisy measurements.

Main Methods:

• Utilizing an image moment-based model for shape estimation, approximating the target as an ellipse.
• Deriving dynamic models for image moments under constant velocity and coordinated turn motions.
• Employing an unscented Kalman filter with an interacting multiple model (UKF-IMM) framework for tracking.
• Developing a likelihood function based on average log-likelihood for the IMM filter.

Main Results:

• The proposed UKF-IMM algorithm successfully estimates the shape of extended objects moving along complex trajectories.
• Simulation results demonstrate the effectiveness of the image moment-based models within the UKF-IMM framework.
• Performance was evaluated using Intersection over Union (IoU) and Root Mean Square Errors (RMSE) for position and velocity.

Conclusions:

• The novel image moment-based approach provides accurate shape estimation and tracking for extended objects.
• The UKF-IMM method effectively handles complex motion patterns and noisy measurement data.
• The proposed algorithm shows competitive performance compared to existing benchmark methods.