Application Effect Analysis for Helmet Detection Algorithm based on YOLOV5
DOI:
https://doi.org/10.54097/hset.v39i.6731Keywords:
Helmet Detection; YOLOV5; Deep Learning.Abstract
Safety helmets can provide better safety protection for the human head in industrial production, which is one of the important protective tools to protect the lives of workers. In the on-site environment, checking whether construction workers are wearing safety helmets mainly depends on manual inspection, which is time-consuming and labor-intensive. In order to solve the problem of hard hat detection and recognition in the complex environment of the construction site, in this paper, we propose a YOLOV5 hard hat detection method. Specifically, a pre-trained object model is used to locate the area of the worker and the face detection module is used to mark the face. Then, the helmet recognition module is used to extract the sub area of the worker's head. Finally, the extracted image is classified into two categories to judge the situation of the worker wearing the helmet. Focusing only on workers can reduce training time and improve the efficiency of the model, so the training label only uses the label of people. The model refines and extracts the head, inputs the separated head area into the subsequent network for subsequent analysis, and then uses the binary classification method to judge whether the helmet is worn. We compare the results of different models such as YOLOV5s, YOLOV5m and YOLOV5l, where YOLOV5l can achieve a best mAP of 91.9%. We also analyze the effect of light to the performance, where the detection effect will be affected by low visibility through the comparative detection of samples in the day and at night.
Downloads
References
Girshick R, Donahue J, Darrell T, et al. Rich featurehierarchies for accurate object detection and semantic segmentation[C]// Proceedings of the 2014 IEEE Conference on Computer Visionand Pattern Recognition. Piscataway: IEEE, 2014: 580-587.
Girshick R. Fast R-CNN[C]// Proceedings of the 2015 IEEEInternational Conference on Computer Vision. Piscataway: IEEE, 2015: 1440-1448.
Ren S Q, He K M, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[J]. arXiv preprint arXiv: 1311. 2524, 2013.
Hao Wu, Jinsong Zhao. An intelligent vision-based approach for helmet identification for work safety[J]. Computers in Industry, 2018, 100: 267-277.
Zhang L L, Lin L, Liang X D, et al. Is faster R-CNN doingwell for pedestrian detection? [C]// Proceedings of the 2016European Conference on Computer Vision, LNCS, 9906. Cham: Springer, 2016: 443-457.
JIao L C, Zhang F, Liu F, et al. A survey of deep learning-based object detection[J]. IEEE Access, 2019, 07: 128837-128868.
Ren S Q, He K M, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (06): 1137-1149.
Liu W, Anguelov D, Erhan D, et al. SSD: sin⁃gle shot multibox detector: European conference on computer vision[C]. Amsterdam: Springer Press, 2016: 21-37.
Redmon J, Divvala S, Girshick R, et al. Youonly look once: unified, real-time object detection: IEEE conference on computer vision & pattern recogni⁃tion[C]. Piscataway NJ: IEEE Press, 2016: 779-788.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
How to Cite
