|
|
基于YOLOv5的电梯行人遮挡检测
|
Abstract:
本文基于YOLOv5对手扶电梯行人的遮挡进行检测，针对电梯场景中行人遮挡导致的检测难的问题，本文在YOLOv5算法的基础上进行了改进。首先，为了解决因遮挡造成的预测框高度重合问题，本文采用了软非极大值抑制(Soft NMS)算法替代传统的NMS算法，该算法通过先对预测框的得分进行衰减处理，然后再进行过滤，从而减少因遮挡导致的人物漏检的情况。其次，针对遮挡导致行人身体可见区域减小、易被忽略的问题，本文在检测端增加了一个微小目标检测头，专门用于检测因遮挡而只部分可见的行人目标，从而提高了模型对这类目标的检测能力。实验结果表明，通过采用以上两种的改进方法对YOLOv5算法进行改进，在电梯行人遮挡检测中能够取得较好的效果。
This article is based on YOLOv5 to detect pedestrian occlusion in escalators. To address the problem of difficulty in detecting pedestrian occlusion in elevator scenes, this article improves the YOLOv5 algorithm. Firstly, in order to solve the problem of high overlap of predicted boxes caused by occlusion, this paper adopts the Soft Non Maximum Suppression (Soft NMS) algorithm instead of the traditional NMS algorithm. This algorithm attenuates the scores of predicted boxes first, and then filters them to reduce the occurrence of missing persons caused by occlusion. Secondly, in response to the problem of occlusion causing a reduction in the visible area of pedestrians and being easily overlooked, this paper adds a small object detection head at the detection end, specifically designed to detect pedestrian targets that are only partially visible due to occlusion, thereby improving the model's detection ability for such targets. The experimental results show that by using the above two improvement methods to improve the YOLOv5 algorithm, good results can be achieved in elevator pedestrian occlusion detection.
| [1] | Boudjit, K. and Ramzan, N. (2021) Human Detection Based on Deep Learning YOLO-v2 for Real-Time UAV Applications. Journal of Experimental & Theoretical Artificial Intelligence, 34, 527-544. https://doi.org/10.1080/0952813x.2021.1907793 |
| [2] | 高英, 吴玉虹. 基于YOLOv5s的轻量化行人检测算法[J]. 现代电子技术, 2023, 46(22): 151-158. |
| [3] | 郭庆梅, 刘宁波, 王中训, 等. 基于深度学习的目标检测算法综述[J]. 探测与控制学报, 2023, 45(6): 10-20+26. |
| [4] | Shao, X., Wei, J., Guo, D., Zheng, R., Nie, X., Wang, G., et al. (2021) Pedestrian Detection Algorithm Based on Improved Faster RCNN. 2021 IEEE 5th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), Chongqing, 12-14 March 2021, 1368-1372. https://doi.org/10.1109/iaeac50856.2021.9390882 |
| [5] | Wang, C., Bochkovskiy, A. and Liao, H.M. (2023) YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, 17-24 June 2023, 7464-7475. https://doi.org/10.1109/cvpr52729.2023.00721 |
| [6] | Chen, Q., Wang, Y., Yang, T., Zhang, X., Cheng, J. and Sun, J. (2021) You Only Look One-Level Feature. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, 20-25 June 2021, 13039-13048. https://doi.org/10.1109/cvpr46437.2021.01284 |
| [7] | Adarsh, P., Rathi, P. and Kumar, M. (2020) YOLO V3-Tiny: Object Detection and Recognition Using One Stage Improved Model. 2020 6th International Conference on Advanced Computing and Communication Systems (ICACCS), Coimbatore, 6-7 March 2020, 687-694. https://doi.org/10.1109/icaccs48705.2020.9074315 |
| [8] | 黄光积. 手扶电梯乘客客流统计和异常行为识别算法研究[D]: [硕士学位论文]. 广州: 华南理工大学, 2023. |
| [9] | 张建军. 基于手扶电梯监控视频的危险行为检测及研究[D]: [硕士学位论文]. 合肥: 安徽大学, 2021. |
| [10] | 陈斌, 陈丽. 基于改进YOLOv5s算法的行人检测方法[J]. 智能物联技术, 2023, 6(3): 34-40. |
| [11] | Kannadasan, B. and Yogeswari, K. (2024) ANN-Based Traffic Planning Model Using Adaptive YOLOv5-Aided Object Counting from Surveillance Videos. Asian Journal of Civil Engineering, 25, 4771-4788. https://doi.org/10.1007/s42107-024-01079-9 |
