Contour coding based rotating adaptive model for human detection and tracking in thermal catadioptric omnidirectional...

Area of Science:

• Computer Vision
• Surveillance Systems
• Thermal Imaging

Background:

• Conventional contour-based methods struggle with TCO vision's distortion.
• Effective human detection and tracking in TCO vision remain challenging.

Purpose of the Study:

• Introduce a novel surveillance system utilizing TCO vision.
• Develop a method to overcome contour-based detection limitations in TCO vision.
• Propose a contour coding based rotating adaptive model (RAM) for feature extraction.

Main Methods:

• Developed a rotating adaptive model (RAM) for contour feature extraction.
• Utilized the relative angle characteristics of TCO vision for adaptive sampling.
• Applied contour coding for direct feature extraction from TCO vision.

Main Results:

• The proposed RAM-based contour coding effectively extracts features from TCO vision.
• Experiments demonstrate satisfactory performance for human detection in TCO vision.
• Quantitative analyses confirm the system's effectiveness for human tracking in TCO vision.

Conclusions:

• The novel RAM-based contour coding is a viable solution for TCO vision challenges.
• The developed surveillance system shows promise for robust human detection and tracking.
• This approach enhances the applicability of TCO vision in surveillance scenarios.