Keeping a pan-tilt-zoom camera calibrated

Area of Science:

• Computer Vision
• Robotics
• Surveillance Technology

Background:

• Pan-tilt-zoom (PTZ) cameras are crucial for surveillance but suffer from coordinate inaccuracies after prolonged use.
• This inaccuracy compromises critical functions like object tracking and 3D localization.

Purpose of the Study:

• To develop a robust calibration model for PTZ cameras that accounts for zoom-dependent focal length and lens distortion.
• To implement a dynamic correction method for maintaining accurate PTZ camera calibration over extended operational periods.

Main Methods:

• A novel PTZ camera model was created, incorporating zoom-dependent focal length and lens distortion.
• Calibration parameters were estimated using nonlinear optimization on a small image set (10 images).
• A dynamic correction algorithm was developed using image feature matching against a stored library.

Main Results:

• The proposed calibration method accurately estimates PTZ camera parameters with minimal images.
• The dynamic correction ensures consistent field of view, even after hundreds of hours of operation.
• Experimental and real-world data validated the effectiveness of both calibration and correction techniques.

Conclusions:

• The developed PTZ camera model and calibration method significantly improve coordinate accuracy.
• The dynamic correction process provides a reliable solution for maintaining PTZ camera performance over time.
• These advancements enhance the reliability of surveillance systems employing PTZ cameras.