YOLOv8n-based personnel detection model for underground mines optimized with SPDs-Conv and WIoU

doi:10.16265/j.cnki.issn1003-3033.2026.05.1332

Abstract

Abstract:

To address the issues of low accuracy and weak robustness in existing detection algorithms due to insufficient lighting, scale differences among personnel, and frequent obstruction by equipment in coal mine environments, as well as the challenges posed by high parameter and computational requirements of some models, which make them difficult to adapt to edge devices underground, an improved YOLOv8n model was proposed to optimize personnel detection tasks in complex mine environments. An enhanced SPDs-Conv module was introduced to enhance the extraction of small target features and improve the recognition accuracy of low-pixel personnel in distant views. Cross stage partial feature fusion + selective kernel attention (C2f_SKAttention) module was designed to strengthen the model's focus on targets of different scales and cope with the scale differences of underground personnel. A dynamic detection head was constructed to adapt to the diversity and complexity of targets, and to improve robustness to occlusion and other scenarios. The WIoU loss function was improved to increase the bounding box localization accuracy and reduce the localization deviation caused by low illumination. The results show that the proposed improved YOLOv8n model achieves an mean average precision (mAP) @0.5 of 83.5% and an mAP@0.5:0.95 of 39.0% on the mine personnel detection dataset. Compared with the original YOLOv8n, the P is improved by 8.5%, the R by 11.9%, the mAP@0.5 by 4.7%, and the mAP@0.5:0.95 by 3.3%. The number of parameters only increases from 3.1M to 3.2M, and the Giga Floating-point operations per second (GFLOPS) rises from 14.0G to 14.4G. The proposed model maintains a lightweight structure while improving detection accuracy and robustness. It effectively alleviates missed detection of small underground targets, insufficient multi-scale adaptation and weak anti-interference capability in complex environments, making it suitable for the limited computing power of underground edge equipment.

Key words: space-to-depth separable convolution (SPDs-Conv), weighted intersection over union (WIoU), YOLOv8n, underground personnel detection, lightweighting, attention mechanism, loss function

CLC Number:

X936

Rong Hai, Xi Zhouyong, Li Jincheng, Pan Xiangyin, Zhang Weida, Han Mingyu. YOLOv8n-based personnel detection model for underground mines optimized with SPDs-Conv and WIoU[J]. China Safety Science Journal, 2026, 36(5): 139-149.

Figures/Tables 10

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Table 1

Table 2

Table 3

Fig.7

References 12

[1]	Diwan T, Anirudh G, Tembhurne J V. Object detection using YOLO: challenges, architectural successors, datasets and applications[J]. Multimedia Tools and Applications, 2023, 82(6): 9243-9275. doi: 10.1007/s11042-022-13644-y
[2]	Rong Hai, He Longyue, Li Nannan, et al. Technical research on surrounding rock control of roadways crossing collapse columns in strong mine pressure working faces[J]. Scientific Reports, 2025, 15(1): DOI:10.1038/s41598-025-15850-8.
[3]	邵小强, 李鑫, 杨永德, 等. 基于改进YOLOv7的矿井人员检测算法[J]. 电子科技大学学报, 2024, 53(3): 414-423.
	Shao Xiaoqiang, Li Xin, Yang Yongde, et al. Mine personnel detection algorithm based on improved YOLOv7[J]. Journal of University of Electronic Science and Technology of China, 2024, 53(3): 414-423.
[4]	贾桂敏, 程羽, 齐孟飞. 多尺度注意力特征增强融合的红外小目标检测新网络[J]. 中国安全科学学报, 2024, 34(6): 90-98. doi: 10.16265/j.cnki.issn1003-3033.2024.06.1565
	Jia Guimin, Cheng Yu, Qi Mengfei. Multi-scale attention feature-enhanced fusion of a new network for infrared small object detection[J]. China Safety Science Journal, 2024, 34(6): 90-98. doi: 10.16265/j.cnki.issn1003-3033.2024.06.1565
[5]	樊红卫, 闫昕山, 曹现刚, 等. 一种DSP-YOLO煤矿带式输送机人员不安全状态智能检测算法[J/OL]. 煤炭科学技术: 1-18. [2025-08-06]. https://link.cnki.net/urlid/11.2402.TD.20250605.1557.008.
	Fan Hongwei, Yan Xinshan, Cao Xiangang, et al. A DSP-YOLO intelligent detection algorithm of unsafe personnel state for coal mine belt conveyor[J/OL]. Coal Science and Technology: 1-18. [2025-08-06]. https://link.cnki.net/urlid/11.2402.TD.20250605.1557.008.
[6]	Alif M A R, Hussain M. YOLOv1 to YOLOv10:a comprehensive review of YOLO variants and their application in the agricultural domain[EB/OL]. arXiv preprint arXiv:2024:DOI:10.48550/arXiv.2406.10139.
[7]	Yang Wenjuan, Wang Yanqun, Zhang Xuhui, et al. MSS-YOLO: multi-scale edge-enhanced lightweight net-work for personnel detection and location in coal mines[J]. Applied Sciences, 2025, 15(6): DOI:10.3390/app15063238.
[8]	Wang Zheng, Hua Zhixin, Wen Yuchen, et al. E-YOLO: recognition of estrus cow based on improved YOLOv8n model[J]. Expert Systems with Applications, 2024, 238:DOI:10.1016/j.eswa.2023.122212.
[9]	王帅, 杨伟, 李宇翔, 等. 基于注意力机制和特征融合的井下轻量级人员检测方法[J]. 煤炭科学技术, 2025, 53(4): 383-392.
	Wang Shuai, Yang Wei, Li Yuxiang, et al. A lightweight personnel detection method for underground coal mines[J]. Coal Science and Technology, 2025, 53(4): 383-392.
[10]	徐飞翔, 张瑞, 程德强, 等. LECODNet: 基于边缘引导与上下文感知的低光环境下无人矿卡遮挡目标检测网络[J/OL]. 煤炭学报: 1-14.[2025-08-06]. https://doi.org/10.13225/j.cnki.jccs.2025.0511.
	Xu Feixiang, Zhang Rui, Cheng Deqiang, et al. LECODNet: edge-guided and context-aware occlusion detection network for autonomous mining trucks under low-light condition[J/OL]. Journal of China Coal Society: 1-14.[2025-08-06]. https://doi.org/10.13225/j.cnki.jccs.2025.0511.
[11]	赵涛, 董立红, 秦昳. 融合注意力和轻量级网络的井下人员检测算法[J/OL]. 计算机工程: 1-10.[2025-08-19]. https://doi.org/10.19678/j.issn.1000-3428.0070383.
	Zhao Tao, Dong Lihong, Qin Yi. Subsurface personnel detection algorithm combining attention and lightweight network[J/OL]. Computer Engineering: 1-10. [2025-08-19]. https://doi.org/10.19678/j.issn.1000-3428.0070383.
[12]	Yue Min, Zhang Liqiang, Huang Juan, et al. Lightweight and efficient tiny-object detection based on improved YOLOv8n for UAV aerial images[J]. Drones, 2024, 8(7): DOI:10.3390/drones8070276.

模块	P/%	R/%	mAP@0.5/%	mAP@0.5:0.95/%	GFLOPs
IoU	86.2	84.8	85.5	83.9	14.2
CIoU	87.5	86.1	86.8	85.3	14.2
SIoU	87.1	85.7	86.4	84.9	14.2
GIoU	86.7	85.3	86.0	84.5	14.2
DIoU	87.3	85.9	86.6	85.1	14.2
WIoUv1	88.8	87.6	88.2	86.8	14.2
WIoUv2	89.5	88.2	88.9	87.5	14.2
WIoUv3	90.3	89.0	89.7	88.3	14.2

SPDs- Conv	尺寸自适应注意力	动态检测头	WIoU	P/%	R/%	mAP@ 0.5/%	mAP@0.5: 0.95/%	参数量/ M	GFLOPs/ G	FPS/ (帧/s)
×	×	×	×	67.6	76.4	78.8	35.7	3.1	14.0	205.6
√	×	×	×	70.2	78.1	80.1	36.5	2.8	14.1	155.1
×	√	×	×	69.5	77.3	79.2	36.2	3.5	14.1	176.5
×	×	√	×	71.8	79.5	80.3	37.1	3.1	14.2	145.2
×	×	×	√	72.3	80.1	81.8	37.5	3.3	14.2	179.4
√	√	×	×	73.5	81.2	81.0	38.0	3.0	14.2	85.1
√	√	√	×	75.2	82.7	81.8	38.7	3.1	14.4	61.5
√	√	√	√	76.1	88.3	83.5	39.0	3.2	14.4	45.2

模型	P/%	R/%	mAP@0.5/%	mAP@0.5:0.95/%	GFLOPs/G	参数量/M	FPS/(帧/s)
YOLOv4	59.5	68.6	70.8	26.2	70.8	64.0	39.5
YOLOv5s	74.5	84.6	62.3	17.6	16.0	7.2	95.2
YOLOv5l	64.2	72.5	70.8	23.9	108.4	44.0	50.6
YOLOv7-tiny	70.5	78.3	65.0	18.5	13.2	6.0	126.1
YOLOv8s	68.1	75.6	69.1	23.6	28.8	11.1	165.2
YOLOv8n	67.6	76.4	78.8	35.7	14.0	3.1	205.6
YOLOv10s	71.5	81.5	79.1	23.5	24.5	8.1	143.0
YOLOv11n	74.9	85.6	73.3	19.4	6.3	2.6	188.1
Ours	76.1	88.3	83.5	39.0	14.4	3.2	45.2

[1]	Cheng Fangming, Hu Jiameng, Gou Rui. A driving risk and steering intention prediction model based on deep learning [J]. China Safety Science Journal, 2026, 36(4): 94-102.
[2]	CAO Haiqing, YAO Zhiying, LYU Shuran, YAO Cuiyou. Evaluation of employee's psychological stress status using LSTM with attention mechanism [J]. China Safety Science Journal, 2026, 36(3): 229-237.
[3]	HU Xiaofeng, TENG Teng, HU Jinming, WEN Jiajun. Identification model of unsafe behaviors among operators in machining workshops [J]. China Safety Science Journal, 2025, 35(8): 40-47.
[4]	WANG Chunyuan, LIU Quanjie. Research on tunnel fire detection based on improved YOLOv8s model [J]. China Safety Science Journal, 2025, 35(3): 69-76.
[5]	XIA Zhenghong, HE Hu, YANG Lei, WU Jianjun, LIU Lu. Aircraft rivet and detachment anomaly detection algorithm based on improved YOLOv8n [J]. China Safety Science Journal, 2025, 35(2): 66-72.
[6]	CAO Shuchao, GE Weibin, LI Conghui, ZHANG Jun. Crowd safety risk assessment based on Seq2Seq-attention network [J]. China Safety Science Journal, 2025, 35(12): 196-203.
[7]	LIU Guangwei, ZHANG Haobo, LEI Jian, HUANG Yunlong, REN Chonghui. Obstacle detection model for unmanned mining trucks during driving in open-pit mines [J]. China Safety Science Journal, 2025, 35(11): 49-55.
[8]	HAO Qinxia, ZHANG Jiaqian. Identification model of miners' unsafe behaviors in coal mine conveyor belt [J]. China Safety Science Journal, 2025, 35(10): 98-105.
[9]	ZHONG Xingrun, TIAN Chenbin, LI Xinhong, MENG Xiaojing, YANG Wenxin. Improved U-Net-based model for urban flood disaster image recognition [J]. China Safety Science Journal, 2025, 35(10): 190-197.
[10]	LIU Huilin, FANG Qiong, JIANG Yu, WEI Huazhang, WANG Tao, ZHANG Shuchuan. A lightweight forest fire detection algorithm based on YOLOv5s [J]. China Safety Science Journal, 2025, 35(1): 75-83.
[11]	LI Gang, DU Yabo, YANG Qinghe, MAO Mengying, JIA Dongping. Obstacle detection on mining roads based on multi-scale feature fusion and attention mechanism [J]. China Safety Science Journal, 2024, 34(9): 87-98.
[12]	JIA Guimin, CHENG Yu, QI Mengfei. Multi-scale attention feature-enhanced fusion of a new network for infrared small object detection [J]. China Safety Science Journal, 2024, 34(6): 90-98.
[13]	QI Lei, LI Ning, LIANG Wei, WANG Zheng, LIU Ziliang. Blade defect detection of U-Net network based on attention mechanism [J]. China Safety Science Journal, 2024, 34(5): 139-146.
[14]	LIU Hong, ZHANG Xindi, LU Fei, ZHANG Chengyu. Research on 4D flight trajectory prediction based on improved Transformer model [J]. China Safety Science Journal, 2024, 34(12): 74-83.
[15]	ZHOU Junchao, CHEN Xin, GAO Jianjie, ZHANG Jie. Urban spatially mixed traffic participants detection model based on improved YOLOv8n [J]. China Safety Science Journal, 2024, 34(12): 178-186.