Aiming at the problems of high cost, difficult implementation and low efficiency of preventing traffic accidents, this paper proposes a deep learning object detection algorithm based on yolov5, which realizes the detection of dangerous driving behavior. Firstly, the dangerous driving data set is labeled and segmented , and re-clustered by K-means method. Secondly, the Non-maximum suppression in yolov5 algorithm is optimized into full category Non-maximum suppression. Finally, the data set is put into the model for iterative training, and the optimal weight is taken. In the test set results, the average accuracy of this algorithm is 97.8%, the average detection time of each image is 8.1ms, and the model size is 13.7mb. Under the same data and experimental environment, this paper compares with the most common dangerous driving behavior detection algorithms such as SSD, yolov3 and yolov4. The results show that this algorithm is better than several common dangerous driving detection algorithms in accuracy, detection speed and model size. In the actual driving process, the algorithm applied in this paper can better meet the requirements of real-time detection and judgment accuracy. Because the model is smaller, the deployment cost in car is lower, and meets the actual requirements.
| Published in | Science Discovery (Volume 10, Issue 4) |
| DOI | 10.11648/j.sd.20221004.16 |
| Page(s) | 242-247 |
| Creative Commons |
This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited. |
| Copyright |
Copyright © The Author(s), 2022. Published by Science Publishing Group |
Dangerous Driving, YOLOV5, Behavior Detection
| [1] | 中华人民共和国国家统计局. 中国统计年鉴[M]. 北京：中国统计出版社, 2019. |
| [2] | 孔令铮. 交通事故致因中的人为因素分析[J]. 中国安全科学学报, 2013, 23 (1): 28-34. |
| [3] | Singh A., Kaur J.. Driver fatigue detection using machine vision approach [C] // IEEE International Advance Computing Conference, 2013. |
| [4] | 赵敏慧．基于图像的驾驶员姿态识别方法研究[D]．东南大学, 2018. |
| [5] | 王坤．基于深度学习的驾驶员分心和疲劳驾驶行为监测方法[D]．武汉理工大学, 2020. |
| [6] | Xiaoxi Ma, Lap-Pui Chau, Kim-Hui Yap. Depth video-based two-stream convolutional neural networks for driver fatigue detection [C] // International Conference on Orange Technologies, 2017. |
| [7] | Xiaoxi Ma, Lap-Pui Chau, Kim-Hui Yap, Guiju Ping. Convolutional Three-Stream Network Fusion for Driver Fatigue Detection from Infrared Videos [C] // IEEE International Symposium on Circuits and Systems, 2019. |
| [8] | Cao G, Xie X, Yang W, et al. Feature-fused SSD: Fast detection for small objects [C]. Ninth International Conference on Graphic and Image Processing (ICGIP 2017). International Society for Optics and Photonics, 2018, 10615: 106151E. |
| [9] | Hurtik P, Molek V, Vlasanek P (2020) YOLO-ASC: you only look once and see contours, accepted. In: Proceedings of IEEE-WCCI conference. |
| [10] | Nan Xiang, Zhao, Cao, Yuedong, Wang, Qianqian Jia．A Real-Time Vehicle TrafficLight Detection Algorithm Based on Modified YOLOv3 [C] // International Conference on Electronics Technology. 2021. |
| [11] | 李昭慧, 张玮良.基于改进YOLOv4算法的疲劳驾驶检测[J].电子测量技术, 2021, 44 (13): 73-78. |
| [12] | 张上, 王恒涛, 冉秀康.基于YOLOv5的轻量化交通标志检测方法[J].电子测量技术, 2022, 45 (08). |
| [13] | Sangdoo Yun, Dongyoon Han, Sanghyuk Chun Seong Joon Oh, Youngjoon Yoo, Junsuk Choe. CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features [C] // International Conference on Computer Vision, 2019. |
| [14] | HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition [C] // 2016 IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778. |
| [15] | ZHENG Z H, WANG P, LIU W, et al. Distance-IoU loss：faster and better learning for bounding box regression [J]. arXiv: 1911.08287, 2019. |
| [16] | Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollar, C. Lawrence Zitnick. Microsoft COCO: Common Objects in Context [C] // European conference on computer vision, 2014. |
APA Style
He Jun, Zhong Kejia, Wu Shengke, Liu Pengzhen. (2022). Research on Dangerous Driving Behavior Detection Based on YOLOV5 Algorithm. Science Discovery, 10(4), 242-247. https://doi.org/10.11648/j.sd.20221004.16
ACS Style
He Jun; Zhong Kejia; Wu Shengke; Liu Pengzhen. Research on Dangerous Driving Behavior Detection Based on YOLOV5 Algorithm. Sci. Discov. 2022, 10(4), 242-247. doi: 10.11648/j.sd.20221004.16
Share This Article
Twitter
Linked In
Facebook
Copy
Download@article{10.11648/j.sd.20221004.16,
author = {He Jun and Zhong Kejia and Wu Shengke and Liu Pengzhen},
title = {Research on Dangerous Driving Behavior Detection Based on YOLOV5 Algorithm},
journal = {Science Discovery},
volume = {10},
number = {4},
pages = {242-247},
doi = {10.11648/j.sd.20221004.16},
url = {https://doi.org/10.11648/j.sd.20221004.16},
eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.sd.20221004.16},
abstract = {Aiming at the problems of high cost, difficult implementation and low efficiency of preventing traffic accidents, this paper proposes a deep learning object detection algorithm based on yolov5, which realizes the detection of dangerous driving behavior. Firstly, the dangerous driving data set is labeled and segmented , and re-clustered by K-means method. Secondly, the Non-maximum suppression in yolov5 algorithm is optimized into full category Non-maximum suppression. Finally, the data set is put into the model for iterative training, and the optimal weight is taken. In the test set results, the average accuracy of this algorithm is 97.8%, the average detection time of each image is 8.1ms, and the model size is 13.7mb. Under the same data and experimental environment, this paper compares with the most common dangerous driving behavior detection algorithms such as SSD, yolov3 and yolov4. The results show that this algorithm is better than several common dangerous driving detection algorithms in accuracy, detection speed and model size. In the actual driving process, the algorithm applied in this paper can better meet the requirements of real-time detection and judgment accuracy. Because the model is smaller, the deployment cost in car is lower, and meets the actual requirements.},
year = {2022}
}
TY - JOUR T1 - Research on Dangerous Driving Behavior Detection Based on YOLOV5 Algorithm AU - He Jun AU - Zhong Kejia AU - Wu Shengke AU - Liu Pengzhen Y1 - 2022/08/17 PY - 2022 N1 - https://doi.org/10.11648/j.sd.20221004.16 DO - 10.11648/j.sd.20221004.16 T2 - Science Discovery JF - Science Discovery JO - Science Discovery SP - 242 EP - 247 PB - Science Publishing Group SN - 2331-0650 UR - https://doi.org/10.11648/j.sd.20221004.16 AB - Aiming at the problems of high cost, difficult implementation and low efficiency of preventing traffic accidents, this paper proposes a deep learning object detection algorithm based on yolov5, which realizes the detection of dangerous driving behavior. Firstly, the dangerous driving data set is labeled and segmented , and re-clustered by K-means method. Secondly, the Non-maximum suppression in yolov5 algorithm is optimized into full category Non-maximum suppression. Finally, the data set is put into the model for iterative training, and the optimal weight is taken. In the test set results, the average accuracy of this algorithm is 97.8%, the average detection time of each image is 8.1ms, and the model size is 13.7mb. Under the same data and experimental environment, this paper compares with the most common dangerous driving behavior detection algorithms such as SSD, yolov3 and yolov4. The results show that this algorithm is better than several common dangerous driving detection algorithms in accuracy, detection speed and model size. In the actual driving process, the algorithm applied in this paper can better meet the requirements of real-time detection and judgment accuracy. Because the model is smaller, the deployment cost in car is lower, and meets the actual requirements. VL - 10 IS - 4 ER -
Email: