首页|基于改进YOLOv5s模型的电梯内电瓶车检测方法

基于改进YOLOv5s模型的电梯内电瓶车检测方法

扫码查看
针对现有检测模型不能满足在实时场景下准确检测电梯内电瓶车的问题,提出一种改进YOLOv5s模型的电梯场景下电瓶车检测方法.改进模型将全局注意力机制融合到YOLOv5s模型的颈部特征融合网络中,来增强模型对融合特征的学习能力;使用充分解耦的检测头部替换原YOLOv5s模型的耦合头部,将目标检测过程中目标分类判定和边框回归这两个子任务解耦,以提升模型检测的准确度.实验结果表明,在自建电梯内电瓶车图像数据集E-Car上,改进YOLOv5s模型显著提高了检测效果,检测指标P、R、mAP_0.5和mAP_0.5:0.95分别达到94.5%、92.2%、96.9%和64.5%,比原模型分别提高4.3%、2.8%、3.8%和4.6%.相比其他主流的目标检测模型,改进YOLOv5 s模型在保持一定检测速度的前提下,具有更高的检测精度,能够实现高效的电梯内电瓶车检测.
Detection Method of Electric Vehicles in Elevator Based on Improved YOLOv5s Model
Aiming at the problem that the existing detection model cannot accurately detect electric vehi-cles in elevators in real-time scenarios,an improved YOLOv5s model is proposed to detect electric vehi-cles in elevator scenarios.The improved model integrates the global attention mechanism into the neck feature fusion network of YOLOv5s model to enhance the model's learning ability of fused features;the fully decoupled detection head is used to replace the coupling head of the original YOLOv5s model,and the target in the detection process is the two subtasks of classification and bounding box regression are decoupled to improve the accuracy of model detection.Experimental results show that on the self-built e-lectric vehicles image data set E-Car in the elevator,the improved YOLOv5s model significantly im-proves the detection effect,and the detection indicators P,R,mAP_0.5 and mAP_0.5:0.95 reach 94. 5%,92.2%,96.9% and 64.5%,are 4.3%,2.8%,3.8% and 4.6% higher than the original model re-spectively.Compared with other mainstream target detection models,the improved YOLOv5s model has a higher detection accuracy while maintaining a certain detection speed,and can realize efficient electric vehicles detection in elevators.

detection of electric vehicles in elevatorYOLOv5sglobal attentionfully decoupled head

李艳秋、缪飞、孙光灵、颜普

展开 >

安徽建筑大学 电子与信息工程学院,安徽 合肥 230601

合肥工业大学 智能互联系统安徽省实验室,安徽 合肥 230009

电梯内电瓶车检测 YOLOv5s 全局注意力 充分解耦头

国家自然科学基金安徽省高等学校协同创新项目合肥工业大学"智能互联系统安徽省实验室"开放基金安徽省住房城乡建设科学技术计划安徽省住房城乡建设科学技术计划

62001004GXXT-2021-024PA2021AKSK01072023-YF0582023-YF113

2024

安徽工程大学学报
安徽工程大学

安徽工程大学学报

影响因子:0.289
ISSN:2095-0977
年,卷(期):2024.39(3)