Continuous Vehicle Detection and Tracking for Non-overlapping Multi-camera Surveillance System
Jinjia Peng, Tianyi Shen, Yafei Wang, Tongtong Zhao, Jun Zhang, Xianping Fu · 2016
Vehicle detection and tracking has always been a significant research on traffic surveillance video. However, multi-camera object tracking consists of a non-overlapping video surveillance network, which makes vehicle re-identification a challenging problem. In this paper, we proposed a novel method for continuous vehicle detection and tracking in multi-camera campus surveillance videos. The method contains two main parts: One is auto vehicle detection and tracking by using background modeling combining with RCNN (Region Convolutional Neural Networks). The other one is multi-camera vehicle re-identification, which collaborates vehicle visual attributes and spatio-temporal information. The experiment results demonstrate that the proposed approach performs with high efficiency and accuracy, which can also be employed to optimize the trajectories of vehicles in multi-camera surveillance videos.