Dynamic camera calibration in support of intelligent transportation systems

An algorithm With which to estimate mean vehicle speed from roadside cameras owned by a traffic management agency is presented. These roadside cameras are not calibrated nor are calibration marks available in the scene. However, estimating camera calibration coefficients is the most important step in extracting quantitative information about the three-dimensional world from the two-dimensional image. Within this framework an algorithm is presented that performs a simplified dynamic calibration and estimates mean vehicle speed. Many algorithms depend on point correspondences between the earth coordinates and the image coordinates as well as targets of known shape to obtain accurate results. However, in the work presented, the goal is to estimate the mean of a distribution of vehicle speeds, and it is demonstrated that a simplified form of calibration is adequate for making an accurate mean speed estimate. Dynamic camera calibration is performed with training sets of 10-s video sequences. The proposed method detects moving vehicles in a set of consecutive frames. This information, together with mean vehicle dimension estimates, is used to create scaling factors that are used to infer a relationship between motion in the image and motion in the earth coordinate system. The proposed algorithm has a camera model with a reduced number of camera calibration parameters. The algorithm is validated with simulated data and actual traffic scenes.