Robust multi-view pedestrian tracking using neural networks

In this paper, we present a real-time robust multi-view pedestrian detection and tracking system for video using neural networks which can be used in dynamic environments. The proposed system consists of two phases: multi-view pedestrian detection and tracking. First, pedestrian detection utilizes background subtraction to segment the foreground objects. An adaptive background subtraction method where each of the pixel of input image models as a mixture of Gaussians and uses an on-line approximation to update the model applies to extract the foreground region. The Gaussian distributions are then evaluated to determine which are most likely to result from a background process. This method produces a steady, real-time tracker in indoor and outdoor environment that consistently deals with changes of lighting condition, and long-term scene change. Second, the tracking is performed at two steps: pedestrian classification and tracking of the individual subject. A sliding window technique is used on foreground binary image which uses for determining the input target patches from input frame. The neural networks is applied for classification with PHOG features of the target patches. Finally, a Kalman filter is applied to calculate the subsequent step for tracking that aims at finding the exact position of pedestrians in an input video frames. The experimental result shows that the proposed approach yields promising performance on multi-view pedestrian detection and tracking on different benchmark datasets.