Robust virtual forces-based camera positioning using a fusion of image content and intrinsic parameters

In this paper, we present a novel and robust model for camera planning in smart vision systems. The proposed approach uses virtual forces to adjust camera parameters (pan and tilt) to the most proper values with respect to the application. The proposed model employs the information in the acquired image and some of the intrinsic camera parameters to estimate pan and tilt displacements required to bring a target object into a specific location of interest in the image. Our model is a general framework and any vision system can be easily modeled to use it. This approach has several advantages over previous work in camera planning. It is portable, expandable, robust, and flexible. Also, there is no need for complicated calibration for the cameras or their pan-tilt heads. The results show that our approach is efficient even with poor system initialization and it is robust against possible weakness in the auxiliary algorithms used.