This paper is concerned with the development and evaluation of the camera calibration method for a real-time head tracking system. Tracking of head movements is important in the design of an eye-controlled human/computer interface and the area of virtual environment.
We propose a video-based head tracking system. A camera is mounted on a subject's head and it takes the front view containing eight 3-dimensional reference points(passive retro-reflecting markers) fixed at the known position(computer monitor). The reference points are captured by image processing board. These points are used to calculate the position (3-dimensional) and orientation of the camera.
A suitable camera calibration method for providing accurate extrinsic camera parameters is proposed. The method have three steps. In the first step, the image center is calibrated using the method of varying focal length. In the second step, the focal length and the scale factor are calibrated from the Direct Linear Transformation(DLT) matrix obtained from the known position and orientation of the camera. In the third step, the position and orientation of the camera is calculated from the DLT matrix, using the calibrated intrinsic camera parameters.
Experimental results show that the average error of camera positions(3-dimensional) is about 0.53˚, the angular errors of camera orientations are less than 0.55˚ and the data aquisition rate is about 10Hz.
The results of this study can be used for the tracking of head movements related to the eye-controlled human/computer interface and the virtual environment.