文献类型：期刊文献

英文题名：A lightweight coal mine pedestrian detector for video surveillance systems with multi-level feature fusion and channel pruning

作者：Xie, Bei Jing Li, Heng Luan, Zheng Li, Xiao Xu Lei, Zhen

第一作者：Xie, Bei Jing

通信作者：Li, H[1]

机构：[1]China Univ Min & Technol Beijing, Sch Emergency Management & Safety Engn, Beijing 100083, Peoples R China;[2]Guizhou Inst Technol, Sch Min Engn, Guiyang 550000, Guizhou, Peoples R China

第一机构：China Univ Min & Technol Beijing, Sch Emergency Management & Safety Engn, Beijing 100083, Peoples R China

通信机构：corresponding author), China Univ Min & Technol Beijing, Sch Emergency Management & Safety Engn, Beijing 100083, Peoples R China.

年份：2025

卷号：15

期号：1

起止页码：5757

外文期刊名：SCIENTIFIC REPORTS

收录：;Scopus(收录号：2-s2.0-85218947395);WOS:【SCI-EXPANDED(收录号:WOS:001424375700037)】；

基金：Much of the work described in this paper was supported by the National Key Research and Development Program of China under Grant No. 2022YFC2904100, and the Central Universities Basic Research Funds Special Funds under Grant No. 2023ZKPYAQ04. The writers are very grateful for the financial support of these projects, expressing their most sincere thanks. We also appreciate the help and guidance provided by Dong Engineer of Guizhou Pannan Coal Development Co., Ltd.in the process of experimental data collection.

语种：英文

外文关键词：Coal mine pedestrian detection; Video surveillance; Lightweight architecture; Channel pruning; Accident prevention

摘要：Pedestrian detection in coal mines is crucial for video surveillance systems. Limited computational resources pose challenges to deploying large models, affecting detection efficiency. To address this, we propose a lightweight pedestrian in coal mine detector with multi-level feature fusion. Our approach integrates the backbone network with coordinate attention, introducing a bidirectional feature pyramid network and a thin neck technique to enhance multi-scale detection capability while reducing computational load. We also employ regression loss with a dynamic focus mechanism for bounding box regression to minimize model errors. The Linkage Channel Pruning method enforces channel-level sparsity on the designed detector to achieve network slimming and secondary lightweight development. Results on a proprietary dataset demonstrate our method's parameters (0.61 M), computational load (2.0 GFLOPs), model size (1.48 MB), detection accuracy (0.966), and inference time (2.1 ms). Compared to the baseline, our method achieves a 4.96 x reduction in parameters, a 4.05 x reduction in computational load, a 4.02 x reduction in model size, a 59.62% reduction in inference time, and a 1.2% accuracy improvement. Experimental validation on proprietary and public datasets confirms that our method exhibits state-of-the-art lightweight performance, accuracy, and real-time capability, demonstrating significant potential in practical engineering applications. The insights gained provide technical references and real-time accident prevention for coal mine video surveillance systems.