PROJECT TITLE :

New Object Detection, Tracking, and Recognition Approaches for Video Surveillance Over Camera Network

ABSTRACT:

Object detection and tracking are 2 fundamental tasks in multicamera surveillance. This paper proposes a framework for achieving these tasks in a very nonoverlapping multiple camera network. A new object detection algorithm using mean shift (MS) segmentation is introduced, and occluded objects are additional separated with the assistance of depth info derived from stereo vision. The detected objects are then tracked by a replacement object tracking algorithm using a novel Bayesian Kalman filter with simplified Gaussian mixture (BKF-SGM). It employs a Gaussian mixture (GM) illustration of the state and noise densities and a novel direct density simplifying algorithm for avoiding the exponential complexity growth of standard Kalman filters (KFs) using GM. When including an improved MS tracker, a replacement BKF-SGM with improved MS algorithm with a lot of sturdy tracking performance is obtained. Furthermore, a nontraining-based object recognition algorithm is used to support object tracking over nonoverlapping network. Experimental results show that: one) the proposed object detection algorithm yields improved segmentation results over standard object detection ways and a pair of) the proposed tracking algorithm can successfully handle advanced situations with smart performance and low arithmetic complexity. Moreover, the performance of both nontraining- and coaching-based mostly object recognition algorithms will be improved using our detection and tracking results as input.