Vehicle Logo Retrieval Based on Hough Transform and Deep Learning
Abstract
Vehicle logo retrieval is an important problem for the intelligent traffic systems, which is still not reliably accurate for practical applications due to the mutable site conditions. In this paper, a new algorithm based on Hough transform and Deep Learning is proposed. The main steps are as follows: First, the logo region is located according to the prior knowledge for the location of vehicle logo and vehicle license plate. Then, typical shapes in vehicle logos, such as circle and ellipse are detected based on optimized Hough transform; meanwhile the accurate position of the logo can be obtained. Finally, the pattern of logo is classified based on Deep Belief Networks (DBNs). Comparative experiments with the actual traffic monitoring images demonstrate that the algorithm outperforms traditional methods in retrieval accuracy and speed. Moreover, the algorithm is particularly suitable for practical application.
Cite
Text
Huan et al. "Vehicle Logo Retrieval Based on Hough Transform and Deep Learning." IEEE/CVF International Conference on Computer Vision Workshops, 2017. doi:10.1109/ICCVW.2017.118Markdown
[Huan et al. "Vehicle Logo Retrieval Based on Hough Transform and Deep Learning." IEEE/CVF International Conference on Computer Vision Workshops, 2017.](https://mlanthology.org/iccvw/2017/huan2017iccvw-vehicle/) doi:10.1109/ICCVW.2017.118BibTeX
@inproceedings{huan2017iccvw-vehicle,
title = {{Vehicle Logo Retrieval Based on Hough Transform and Deep Learning}},
author = {Huan, Li and Li, Wang and Qin, Yujian},
booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
year = {2017},
pages = {967-973},
doi = {10.1109/ICCVW.2017.118},
url = {https://mlanthology.org/iccvw/2017/huan2017iccvw-vehicle/}
}