Abstract:
An automatic method is proposed to calibrate an outdoor PTZ (pan-tilt-zoom) camera system with high precision, which is successfully applied to human height measurement in visual monitoring systems. Specifically, the whole camera is divided into a zoom transformation model to deal with the changing zoom parameters, and a fixed focal camera parameters model. For the former, a calibration method based on multiple reference points is presented. For the latter, rotation angle information of the camera is utilized to conduct on-line calibration for the fixed focal camera. For the reason that both the zoom center and the principal point are considered as independent variables which are calibrated in zoom transformation model and fixed focal camera model respectively, the proposed model can achieve high accuracy. Furthermore, since multiple reference points are adopted to guarantee that the modeling error is small and there exist sufficient matching features even when the zooming range is large, high calibration precision is achieved. Based on the self-calibration results, a height measurement approach for the human in monitoring environments is proposed. Experimental results are provided to demonstrate the performance of both the self-calibration and height measurement algorithms.