Abstract
The accuracy of the non-overlapping camera network topology estimation can improve the accuracy of cooperative tracking across non-overlapping cameras. Aiming at the problem of low precision of travel time estimation in camera network topology estimation, we propose a camera network topology estimation based on blind distance. Firstly, the average velocity is obtained by tracking the target motion trajectory in the field of view of the camera. Secondly, the arrival time is estimated by the re-identification across non-overlapping cameras. Finally, a Gaussian distribution between the cameras is estimated by using the cross-correlation function. The effectiveness of our approach is validated on the MCT dataset. Our approach outperforms the state-of-the-art camera network topology estimation.