基于改进YOLOv8的低光照行人检测算法  

Low Light Pedestrian Detection Algorithm Based on Improved YOLOv8

作　　者：李宝兵 符长友 LI Baobing;FU Changyou(School of Computer Science and Engineering,Sichuan University of Science and Engineering,Zigong 643000 China)

机构地区：[1]四川轻化工大学计算机科学与工程学院,四川自贡643000

出　　处：《西华大学学报(自然科学版)》2025年第3期29-36,共8页Journal of Xihua University:Natural Science Edition

基　　金：企业信息化与物联网测控技术四川省高校重点实验室项目(2021WYY02)。

摘　　要：行人检测被广泛应用于智能交通和自动驾驶领域。在低光照场景下,行人检测存在漏检、误检等问题,检测精度低。为此,提出一种基于改进YOLOv8n的低光照行人检测(GSGYOLOv8)算法:首先,在主干网络添加GCNet模块,提高模型对图像上下文信息的提取能力;然后,在主干网络融合SPDConv和Conv,增强模型对局部特征的提取能力,提升小目标检测的效果;最后,在颈部网络添加GAM注意力机制,自适应地调整目标与背景之间的相关程度,降低背景信息的干扰。相较于基线YOLOv8n算法,该算法在NightSurveillance数据集上mAP@0.5和mAP@0.5~0.95分别提升了3.4和4.6;相较于其他主流算法,该算法系统开销更低,目标检测精度更高,特别是克服了低光照的影响,提升了低光照条件下行人检测的精度。Pedestrian detection is widely used in the fields of intelligent transportation and autonom-ous driving.However,in low light scenarios,pedestrian detection has problems such as missed detection and false detection,which result in reduced detection accuracy.Therefore,a low light pedestrian detection algorithm GSG-YOLOv8 based on YOLOv8 improvement is proposed.Firstly,the GCNet module is ad-ded to the backbone network to enhance the model's ability to extract contextual information from images.Then,SPDConv and Conv are integrated into the backbone network to enhance the model's ability to ex-tract local features and improve the effectiveness of small object detection.Finally,GAM attention mechan-ism is added to the neck network to adaptively adjust the correlation between the target and background,and reduce the interference of background information.Compared to the baseline YOLOv8n algorithm,the improved algorithm performs mAP@0.5 And mAP@0.5~0.95 increased by 3.4 and 4.6 respectively.Com-pared to other mainstream algorithms,the improved algorithm has lower system overhead and higher ob-ject detection accuracy.The experimental results show that the GSG-YOLOv8 algorithm overcomes the in-fluence of low light and improves the accuracy of pedestrian detection under low light conditions.

关 键 词：行人检测 YOLOv8 低光照 特征提取 小目标检测 注意力机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]