Behavioral Analysis and Individual Tracking Based on Kalman Filter: Application in an Urban Environment

Area of Science:

• Computer Vision
• Artificial Intelligence
• Urban Surveillance Systems

Background:

• Improving behavioral analysis in urban environments is crucial for safety and management.
• Existing tracking methods struggle with occlusions and individuals crossing paths in video data.

Purpose of the Study:

• To propose and evaluate a novel people tracking method for urban video surveillance.
• To compare the efficacy of a proximity-based tracking approach with a Kalman filter-based method.

Main Methods:

• Developed two tracking approaches: proximity-based comparison of positions between frames.
• Implemented a Kalman filter-based method for predicting individual positions in subsequent images.
• Integrated distance concepts from the proximity method into the Kalman filter approach.

Main Results:

• Kalman filter tracking proved more robust, successfully handling occlusions and person crossings.
• The enhanced Kalman filter method improved both tracking accuracy and abnormal behavior detection.
• Experimental results confirmed the Kalman filter method's superiority over the proximity method alone.

Conclusions:

• Kalman filters offer a reliable approach for people tracking in complex urban surveillance scenarios.
• The integration of distance metrics enhances Kalman filter performance for improved behavioral analysis.
• The proposed methods provide valuable insights into speed and trajectory variations for urban monitoring.