Auto Camera Calibration Method for Person Tracking Applications Bruno Merven, Fred Nicolls, and Gerhard de Jager Department of Electrical Engineering, University of Cape Town Private Bag, Rondebosch, 7701, South Africa {bruno,nicolls,gdj}@dip.ee.uct.ac.za Abstract. Jones [1] recently presented a novel calibration procedure that uses a linearized model of the projection of the height of a person to recover the image-plane to local-ground- plane transformation with minimum expert intervention. In this paper we present an addition to the work of Jones [1], by further reducing the measurements needed to be made manually to de ne the image-plane to ground-plane homography. We also do an analysis of the sensitivity of the recovered transformation to errors in measurements and noise, and point out how to use the method to achieve best calibration results. Keywords: Auto camera calibration, person tracking 1 Introduction A good camera model for image-plane to ground-plane transformations is a useful thing to have when tracking people or moving objects within monitored scenes. It allows the tracking to take place in world coordinates rather than in image coordinates and this has several advantages: 1. Motion models are easier to construct in world coordinates; 2. Constraints on motion are more easily imposed; 3. Occlusions become easier to ressolve; 4. The definition of a common coordinate system in the case of multi-camera tracking configurations is made simpler. The camera model as described by Tsai [2] is based on the pin hole model of perspective projection. This model is defined by a set of intrinsic parameters and extrinsic parameters. Intrinsic parameters are the internal properties of the camera that describe how the camera forms an image. Tsai [2] includes the following internal param- eters: focal length, pixel dimensions, distortion coefficients and the pixel coordi- nates of the optical center. Extrinsic parameters describe the camera’s pose: orientation and position in the world coordinate frame. The external parameters are simply the rotation J. Bigun and T. Gustavsson (Eds.): SCIA 2003, LNCS 2749, pp. 91–100, 2003. c  Springer-Verlag Berlin Heidelberg 2003