A semi-automatic motion-constrained Graph Cut algorithm for Pedestrian Detection in thermal surveillance videos

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Pedestrian detection is crucial for safety systems.
• Thermal infrared imaging offers advantages in various lighting conditions.
• Existing methods may lack robustness or generalizability.

Purpose of the Study:

• To develop a semi-automatic algorithm for robust pedestrian detection in thermal infrared images.
• To improve the accuracy and reliability of pedestrian detection using Graph Cut optimization.
• To create a generalizable algorithm effective across diverse datasets.

Main Methods:

• A semi-automatic algorithm leveraging Graph Cut optimization for binary labeling.
• Incorporation of an additional term in the energy formulation to bias detection towards pedestrians.
• Utilizing user-selected seeds and motion constraints for enhanced reliability.

Main Results:

• The algorithm achieved an average precision of 98.92% and an average recall of 99.25% across four public databases.
• Demonstrated superior performance compared to several existing methods and state-of-the-art approaches on the same datasets.
• The method showed good generalization capabilities across different databases.

Conclusions:

• The proposed semi-automatic algorithm provides a reliable and robust solution for pedestrian detection in thermal infrared imagery.
• The integration of Graph Cut optimization with specific biasing terms and constraints yields state-of-the-art results.
• The algorithm's effectiveness and generalizability make it a valuable tool for various applications.