Education, Science, Technology, Innovation and Life
Open Access
Sign In

Pedestrian and vehicle detection method in infrared scene based on improved YOLOv5s model

Download as PDF

DOI: 10.23977/autml.2024.050111 | Downloads: 0 | Views: 45

Author(s)

Jie Yang 1, Wenzhun Huang 1

Affiliation(s)

1 School of Electroinc Information, Xijing University, Xi'an, China

Corresponding Author

Wenzhun Huang

ABSTRACT

A infrared pedestrian and vehicle target detection method based on an improved YOLOv5s model is proposed to address the issues of false alarms and missed detections caused by small pedestrian and vehicle targets, occlusion, and low visibility in nighttime driving and complex environments. To address the issue of missed detection of small targets, a small target detection layer is introduced, which increases the three detection layers of the original model to four layers to better handle the detection problem of small-sized targets; The SIoU loss function has been introduced to improve the accuracy of multi-scale object detection, allowing the model to better process different types of targets separately, enhancing the flexibility and generalization ability of the model; At the same time, in response to the contradiction between different tasks of the model, which leads to low pedestrian and vehicle detection accuracy and slow convergence speed, a decoupling head is introduced in the YOLOv5s head to improve the model detection accuracy and positioning speed; Experiments were conducted on the FLIR dataset, and the results showed that the improved YOLOv5s model algorithm increased Precision by 1.6% compared to the original YOLOv5s model algorithm, with Recall and mAP @ 0 5 and mAP @ 0 5: 0 95% has increased by 3.4%, 2.3%, and 5.2%, reaching 90.0%, 88.2%, 93.9%, and 58.9%, respectively.

KEYWORDS

Object detection, Infrared scene, YOLOv5s model, Small goals, Obscure the target, SIoU loss function, Decoupling head

CITE THIS PAPER

Jie Yang, Wenzhun Huang, Pedestrian and vehicle detection method in infrared scene based on improved YOLOv5s model. Automation and Machine Learning (2024) Vol. 5: 90-96. DOI: http://dx.doi.org/10.23977/autml.2024.050111.

REFERENCES

[1] Jiaxiang Q, Ziming W, Yimin H.An embedded device-oriented fatigue driving detection method based on a YOLOv5s [J].Neural Computing and Applications, 2023, 36(7):3711-3723.
[2] Yu C, Shin Y.SAR ship detection based on improved YOLOv5 and BiFPN [J].ICT Express, 2024, 10(1):28-33.
[3] Jun X, Renjie G, Yuanpei Z, et al.Carbonate Rock Fracture Identification Method Based on an Improved YOLOv5 Algorithm[J].Pure and Applied Geophysics, 2024, 181(1):189-201.
[4] Kyedong L, Sik K P.Deep Learning Model Analysis of Drone Images for Unauthorized Occupancy Detection of River Site [J].Journal of Coastal Research, 2024, 116(sp1):284-288.
[5] Xing B, Sun M, Ding M, et al.Fish sonar image recognition algorithm based on improved YOLOv5.[J].Mathematical biosciences and engineering : MBE, 2024, 21(1):1321-1341.
[6] L. L M, L. Z, X. C, et al.Research on surface defect detection method of metallurgical saw blade based on YOLOV5[J].Metalurgija, 2024, 63(1):121-124.
[7] Yanru F, Yuliang C, Huijun Y.A detection algorithm based on improved YOLOv5 for coarse-fine variety fruits [J].Journal of Food Measurement and Characterization, 2023, 18(2):1338-1354.
[8] Ziwei W, Yi H, Jianxiang D, et al. YOLOv5-Based Seabed Sediment Recognition Method for Side-Scan Sonar Imagery [J]. Journal of Ocean University of China, 2023, 22(6):1529-1540.
[9] Luxuan B, Bo L, Jue W, et al.Multi-branch stacking remote sensing image target detection based on YOLOv5[J].The Egyptian Journal of Remote Sensing and Space Sciences, 2023, 26(4):999-1008.
[10] Yunfeng J, Zhizhan L, Ruili W.Research on lightweight pedestrian detection based on improved YOLOv5[J]. Mathematical Models in Engineering, 2023, 9(4):178-187.

Downloads: 1658
Visits: 69476

Sponsors, Associates, and Links


All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.