机电工程技术2025，Vol.54Issue(6)：78-83,6.DOI:10.3969/j.issn.1009-9492.2025.06.014
基于改进YOLOv8的丁烷气体红外图像检测算法
Infrared Image Detection Algorithm for Butane Gas Based on Improved YOLOv8
摘要
Abstract
An improved YOLOv8 algorithm is proposed to address the issues of low accuracy,slow speed,and susceptibility to environmental influences in existing infrared imaging based butane gas leak detection methods.On the basis of retaining the high-precision detection advantage of YOLOv8,the detection performance is significantly improved through three aspects of optimization.Firstly,the use of deformable convolution technology instead of traditional convolution in YOLOv8 enhances the model's adaptability to the shape variability of leaked gases.Secondly,embedding an adaptive multi head attention mechanism module in the Neck network effectively improves the model's ability to extract and recognize features with unclear contours and low contrast.Finally,the model's generalization ability and detection accuracy are improved by the introduction of the WIoU loss function.The experimental results show that on the self built dataset,the algorithm achieves a detection accuracy of 87.2%and mAP@0.5 of 89.7%.The average number of detected image frames per second reaches 7.6,which outperforms commonly uses algorithms in terms of performance.It improves algorithm not only enhances detection speed but also ensures high accuracy,providing an efficient and reliable solution for butane gas leak detection.It is expected to play an important role in practical applications and provide strong guarantees for industrial safety.关键词
丁烷/红外图像/YOLOv8/可变形卷积/自适应多头注意力机制/WIoUKey words
butane/infrared image/YOLOv8/deformable convolution/adaptive multiple attention mechanism/WIoU分类
信息技术与安全科学引用本文复制引用
于航,郭家乐,张佳文,夏光庆,张龙刚..基于改进YOLOv8的丁烷气体红外图像检测算法[J].机电工程技术,2025,54(6):78-83,6.基金项目
国家自然科学基金(61774130) （61774130）
云南师范大学研究生科研创新基金项目(YJSJJ23-B112) （YJSJJ23-B112）
