基于改进YOLOX轻量级的烟雾火焰目标检测方法  

Improved YOLOX-based Lightweight Fire Object Detection Algorithm

作　　者：金程拓 赵永智 王涛[3] 刘惠临[1] Jin Chengtuo;Zhao Yongzhi;Wang Tao;Liu Huilin

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽淮南232001 [2]国家能源集团宁夏煤业公司应急救援中心,宁夏灵武751400 [3]滁州学院无人应急装备与灾害客数字化重建安徽省联合共建学科重点实验室,安徽滁州239000

出　　处：《滁州学院学报》2023年第5期40-45,共6页Journal of Chuzhou University

基　　金：安徽省质量工程项目现代产业学院(2021cyxy054);安徽省高等学校科研计划项目“留守儿童溺水突发事件敏捷预警建模与响应技术研究”(2022AH051101)。

摘　　要：本文针对目前基于深度学习的烟雾火焰目标检测任务中存在的在嵌入式设备下准确率低、实时性能差的问题,提出一种基于改进的轻量级YOLOX算法的烟雾火焰目标检测方法F-YOLOX。该方法以YOLOX-s作为基准模型,通过引入ECA注意力模块、使用SIoU损失函数和采用GELU激活函数等技术手段对原模型进行改进。实验结果表明,F-YOLOX在烟焰图像检测数据集上检测平均精度达到78.45%,比基准模型提高3.7%;检测速度达到53.4fps,比基准模型提高2.8%。F-YOLOX能够满足烟雾火焰检测任务精度和实时性的要求。A smoke and flame detection method based on an improved lightweight YOLOX algorithm F-YOLOX is proposed to address the problems of low accuracy and poor real-time performance in smoke and flame object detection tasks using deep learning on embedded devices.The method uses YOLOX-s as the baseline model and improves it by introducing the ECA attention module,using the SIoU loss function,and adopting the GELU activation function.The experimental results demonstrate that F-YOLOX achieves an average precision of 78.45%on the flame image detection dataset,which is an improvement of 3.7%compared to the baseline model.Additionally,F-YOLOX achieves a detection speed of 53.4 fps,which is a 2.8%improvement over the baseline model.Therefore,F-YOLOX is capable of meeting the requirements for both accuracy and real-time performance in smoke and flame detection tasks.

关 键 词：目标检测 视频烟雾火焰检测 深度学习 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]