Real-Time personnel safety detection in chemical parks using YOLOv8-ARR

• Published in: AIMS electronics and electrical engineering Vol. 9; no. 3; pp. 260 - 287
• Main Authors: Wang, Zhong, Lei, Lanfang, Li, Tong, Shi, Peibei
• Format: Journal Article
• Language: English
• Published: AIMS Press 01.09.2025
• Subjects: attention mechanism rcpca; chemical industrial park; real-time monitoring system; safety behavior detection; yolov8-arr algorithm
• ISSN: 2578-1588; 2578-1588
• DOI: 10.3934/electreng.2025013

With the rapid development of the chemical industry, personnel safety has become a critical concern. Traditional monitoring systems, which rely heavily on human surveillance, are inefficient and often inaccurate. To address this issue, we proposed a real-time detection and identification system for personnel safety in chemical parks based on an advanced artificial intelligence algorithm named YOLOv8-ARR. The key contributions of this system include: (1) The introduction of Adaptive Powerful-IoU(APIoU) as a network optimization for bounding box regression loss, which effectively balances gradient gains between high-quality and low-quality samples, enhancing model localization; (2) a novel attention mechanism, Reinforced Channel Prioritized Contextual Attention(RCPCA), to improve background information extraction; (3) replacing traditional convolution with RFAConv to assign different weights to each receptive field position and feature channel, highlighting crucial details; (4) the use of a bidirectional feature pyramid network (BiFPN) for the weighted fusion of multi-scale feature maps; and (5) the addition of a small object detection layer in the YOLOv8 network to enhance the detection of small targets. Experimental results on a custom dataset of chemical park workers showed that our model improves the mean Average Precision (mAP@0.5) by 5.475% compared to the standard model. This system provides a more accurate solution for identifying abnormal behaviors and potential risks in chemical parks compared to traditional methods. Additionally, it significantly reduces dependency on human resources, minimizes false positives and negatives, and enhances monitoring efficiency and safety.