A fuzzy decision-making system for video tracking with multiple objects in non-stationary conditions

Area of Science:

• Computer Vision
• Robotics
• Artificial Intelligence

Background:

• Tracking multiple moving objects in dynamic environments is a persistent challenge in computer vision.
• Existing methods struggle with non-stationary settings, impacting applications like surveillance and autonomous navigation.

Purpose of the Study:

• To develop and evaluate a novel computer vision algorithm for robust multi-object tracking in non-stationary environments.
• To enhance the capabilities of autonomous navigation systems through accurate real-time object monitoring.

Main Methods:

• Utilizes the Kanade-Lucas-Tomasi (KLT) feature tracker to identify and track feature points between successive frames.
• Employs movement information and background subtraction to isolate moving objects.
• Applies fuzzy logic, using mass center and length-to-width ratio, for object categorization and segregation.

Main Results:

• The algorithm successfully tracks and classifies vehicles, pedestrians, bicycles, and motorcycles with approximately 75% accuracy.
• Achieves a processing speed of 43 frames per second, demonstrating superior performance over existing approaches.
• Effectively monitors moving objects in real-time within the camera's field of view.

Conclusions:

• The proposed method offers a significant advancement in multi-object tracking for autonomous navigation and surveillance.
• The algorithm provides a balance of high accuracy and efficient processing speed, making it suitable for real-time applications.
• Fuzzy logic integration enables effective categorization of diverse moving objects.