基于改进YOLOv8的电梯轿厢禁入目标检测算法研究  

Research on Detection Algorithm of Prohibited Objects in Elevator Car Based on Improved YOLOv8

作　　者：程楠 任永强 李贵霖 黄子麟 冯金奎 Cheng Nan;Ren Yongqiang;Li Guilin;Huang Zilin;Feng Jinkui(Key Laboratory of Special Equipment Safety and Energy-saving,State administration for Market Regulation,Beijing 100029;China Special Equipment Inspection&Research Institute,Beijing 100029)

机构地区：[1]国家市场监督管理总局重点实验室(特种设备安全与节能),北京100029 [2]中国特种设备检测研究院,北京100029

出　　处：《中国特种设备安全》2025年第2期16-21,共6页China Special Equipment Safety

基　　金：北京市自然科学基金资助项目(3234062);国家市场监督管理总局科技计划项目(2022MK204);中国特检院内部项目(2021青年20)。

摘　　要：电梯轿厢内禁入目标入户易造成火灾风险。针对电梯轿厢视频监控下,现有视觉检测方法存在禁入目标误检和漏检的问题,提出了一种基于改进YOLOv8的电梯轿厢禁入目标检测算法。首先在数据预处理阶段对禁入目标检测数据集进行数据增强,然后对模型引入可变形注意力模块,最后将原始YOLOv8的损失函数替换为SIou函数。实验结果表明,相比于YOLOv8基准模型,本文方法的评价指标值具有明显提升,表明该方法能够有效检测电梯轿厢禁入目标。The entry of prohibited objects into the elevator car may easily cause fire risks.Under the video surveillance of the elevator car,the existing visual detection method has the problem of misdetection and missed detection of prohibited objects.An improved YOLOv8-based algorithm for detecting prohibited objects in elevator cars is proposed.Firstly,in the stage of data preprocessing,data enhancement is performed on the prohibited object detection dataset.Secondly,the deformable attention module is introduced into the model.Finally,the loss function of the original YOLOv8 is replaced with SIoU.The experimental results show that compared with the YOLOv8 benchmark model,the evaluation indicators of this method are significantly improved,indicating that the proposed method can effectively detect prohibited objects in elevator cars.

关 键 词：YOLOv8 电梯轿厢 禁入目标检测 可变形注意力 SIou函数 

分 类 号：X941[环境科学与工程—安全科学]