Object Detection and Tracking with YOLO and the Sliding Innovation Filter

Area of Science:

• Computer Vision
• Machine Learning
• Artificial Intelligence

Background:

• Object detection and tracking are crucial in machine learning and computer vision.
• Existing methods struggle with real-world challenges like occlusions and disturbances.

Purpose of the Study:

• To propose a novel object detection and tracking approach.
• To address challenges in dynamic and uncertain environments.

Main Methods:

• Integration of You Only Look Once (YOLO) for object detection.
• Implementation of a sliding innovation filter for robust target tracking.
• Estimation of optimal centroid location and trajectory updating.

Main Results:

• The proposed sliding innovation filter-based tracking outperforms traditional Kalman-based methods.
• Enhanced tracking reliability in the presence of disturbances was demonstrated.
• Experimental simulations in surveillance scenarios validated the approach.

Conclusions:

• The research offers a practical and effective solution for object detection and tracking.
• The sliding innovation filter proves robust in disturbed and uncertain environments.
• This work provides a foundation for advancing multi-object tracking applications.