The Counting App, or How to Count Vehicles in 500 Hours of Video
Abstract
This paper proposes a new method for counting vehicles based on video tracking. The process consists of two main steps: tracking vehicles and processing the output with minimal user input, separating the vehicle positions into sets of trajectories, which correspond to the paths drivers can take. The method allows to rapidly analyze videos from road sections and intersections, and yields detailed results in the form of turning movement counts. A large dataset of five hundred hours of traffic videos was processed using this method and the results are promising as mean absolute percentage error (MAPE) can get as low as 14 % depending on the conditions and the quality of the video capture. This paper also discusses the factors that affect counting performance and how to improve counting accuracy.
Cite
Text
Lessard et al. "The Counting App, or How to Count Vehicles in 500 Hours of Video." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2016. doi:10.1109/CVPRW.2016.198Markdown
[Lessard et al. "The Counting App, or How to Count Vehicles in 500 Hours of Video." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2016.](https://mlanthology.org/cvprw/2016/lessard2016cvprw-counting/) doi:10.1109/CVPRW.2016.198BibTeX
@inproceedings{lessard2016cvprw-counting,
title = {{The Counting App, or How to Count Vehicles in 500 Hours of Video}},
author = {Lessard, Adrien and Bélisle, François and Bilodeau, Guillaume-Alexandre and Saunier, Nicolas},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2016},
pages = {1592-1600},
doi = {10.1109/CVPRW.2016.198},
url = {https://mlanthology.org/cvprw/2016/lessard2016cvprw-counting/}
}