Vehicle Detection Algorithm Based on Fisheye Camera in Parking Environment

Liao Wang, Qiu Fang

2023

Abstract

To address the issue of missing and wrong detection of fisheye images by existing target detection algorithms, a targeted dataset is constructed, and an improved model is proposed using YOLOv5 as a reference. Firstly, to facilitate better adapt to the dimensions of the custom dataset, the K-means++ algorithm was utilized for anchor box clustering. Secondly, a deformable convolutional network was brought in to substitute some convolution layers in the original network, so that the network can adaptively extract distorted image feature points, and the integration of coordinate attention mechanisms enhances the expression of semantic and positional information of the feature points of interest. Furthermore, Slim-Neck is designed to replace the original Neck based on GSConv convolution, resulting in reduced model parameters and enhanced algorithmic precision rate. Lastly, redesigning the detector's loss function by incorporating EIoU Loss and Focal Loss. The results demonstrate that precision rate, recall rate and mean precision are improved by 2.31%, 4.41% and 3.50% respectively compared with YOLOv5 algorithm.

Paper Citation

in Harvard Style

Wang L. and Fang Q. (2023). Vehicle Detection Algorithm Based on Fisheye Camera in Parking Environment. In Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT; ISBN 978-989-758-677-4, SciTePress, pages 5-10. DOI: 10.5220/0012272600003807

in Bibtex Style

@conference{anit23,
author={Liao Wang and Qiu Fang},
title={Vehicle Detection Algorithm Based on Fisheye Camera in Parking Environment},
booktitle={Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT},
year={2023},
pages={5-10},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012272600003807},
isbn={978-989-758-677-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT
TI - Vehicle Detection Algorithm Based on Fisheye Camera in Parking Environment
SN - 978-989-758-677-4
AU - Wang L.
AU - Fang Q.
PY - 2023
SP - 5
EP - 10
DO - 10.5220/0012272600003807
PB - SciTePress