Vehicle Re-Identifiation and Multi-Camera Tracking in Challenging City-Scale Environment

Jakub Spanhel, Vojtech Bartl, Roman Juranek, Adam Herout; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 150-158

Abstract

In our submission to the NVIDIA AI City Challenge, we address vehicle re-identification and vehicle multi-camera tracking. Our approach to vehicle re-identification is based on the extraction of visual features and aggregation of these features in the temporal domain to obtain a single feature descriptor for the whole observed track. For multi-camera tracking, we proposed a method for matching vehicles by the position of trajectory points in real-world space (linear coordinate system). Furthermore, we use CNN for vehicle re-identification task to filter out false matches generated by proposed positional matching method for better results.

Related Material

[pdf]
[bibtex]
@InProceedings{Spanhel_2019_CVPR_Workshops,
author = {Spanhel, Jakub and Bartl, Vojtech and Juranek, Roman and Herout, Adam},
title = {Vehicle Re-Identifiation and Multi-Camera Tracking in Challenging City-Scale Environment},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}
}