Pedestrian Tracking Using Single Camera with New Extended Kalman Filter

In this paper, a new system is developed for pedestrian counting and tracking in the scene under occlusion. The camera is installed in a straight and stationary position. Main features of the proposed system are tracking under different light conditions (day and night) and Pose estimation using a new Extended Kalman filter (EKF). The system outputs the spatio-temporal coordinates of each pedestrian during the period he/she is in the scene. The system has three main parts: blobs detection, increasing shadow for prediction of positions and using the proposed EKF for the tracking purpose. Blob detection is performed using the dynamic background difference method. The basic idea of the proposed EKF is to linearize the state-space model at each time instant in the High Dimensional Space (HDS). The HDS helps to have a linear space from nonlinear space. In this space, standard Kalman filter can apply for getting best results in estimation and prediction procedures. It is proven that MSE and error variance in this space is less than input space. Experimental results are based on computer simulation, indoor and outdoor scenes and demonstrate the system’s robustness under partial or full occlusions of pedestrians.

کلیدواژه ها:

Pedestrian tracking ، dynamic background ، extended Kalman filter ، high dimensional space.