Vehicle Re-Identification for Automatic Video Traffic Surveillance [ATS-CVPR 2016]

Dominik Zapletal and Adam Herout
GRAPH@FIT, Brno University of Technology
Corresponding authors: herout [at] fit.vutbr.cz

Abstract

This paper proposes an approach to the vehicle reidentification problem in a multiple camera system. We focused on the re-identification itself assuming that the vehicle detection problem is already solved including extraction of a full-fledged 3D bounding box. The re-identification problem is solved by using color histograms and histograms of oriented gradients by a linear regressor. The features are used in separate models in order to get the best results in the shortest CPU computation time. The proposed method works with a high accuracy (60 % true positives retrieved with 10 % false positive rate on a challenging subset of the test data) in 85 milliseconds of the CPU (Core i7) computation time per one vehicle re-identification assuming the fullHD resolution video input. The applications of this work include finding important parameters such as travel time, traffic flow, or traffic information in a distributed traffic surveillance and monitoring system.

Downloads

  • Paper (OpenAccess CV-F repository)
  • VehicleReId – dataset used for training and evaluation of the algorithm in the paper.
    • Dataset details can be found in the paper
    • See README in the zip file for further information about the dataset structure
    • Some statistics about the dataset:
      • 5 video shots taken from two cameras
      • 47,123 extracted vehicle images
      • 24,530 human anotated vehicle pairs

Citation

@INPROCEEDINGS{Zapletal2016,
 author = {Zapletal, Dominik and Herout, Adam},
 title = {Vehicle Re-Identification for Automatic Video Traffic Surveillance},
 pages = {1--7},
 booktitle = {International Workshop on Automatic Traffic Surveillance (CVPR 2016)},
 year = {2016},
 location = {Las Vegas, US},
 publisher = {IEEE Computer Society},
 ISBN = {978-0-7695-4989-7},
 language = {english}
}

VehicleReId_teaser

Acknowledgements

This work was partially supported by TACR grant TE01020415 “V3C”, by TACR project “RODOS”, TE01020155, and by The Ministry of Education, Youth and Sports of the Czech Republic from the National Programme of Sustainability (NPU II); project IT4Innovations excellence in science -LQ1602.