关于| 联系| 地图| 邮箱
首页>>>学术期刊>>>期刊年表>>>2024年论文集(45卷)
2024年论文集(45卷) 2023年论文集(44卷) 2022年论文集(43卷) 2021年总目录(42卷) 2020年总目录(41卷) 2019年总目录(40卷) 2018年总目录(39卷) 2017全年目录(38卷) 2016全年目录(37卷) 2015全年目录(36卷) 2014全年目录(35卷) 2013全年目录(34卷) 2012全年目录(33卷) 2011全年目录(32卷) 2010全年目录(31卷) 2009全年目录(30卷) 2008全年目录(29卷) 2007全年目录(28卷) 2006全年目录(27卷) 2005全年目录(26卷) 2004全年目录(25卷) 2003全年目录(24卷) 2002全年目录(22卷) 2001全年目录(22卷) 2000全年目录(21卷)

20240206基于MBE-YOLOv5的轻量化化工袋目标检测方法

‖  文章供稿:刘伟鑫 林邦演 黄汉亿 李旻龙
‖  字体: [大] [中] [小]

刘伟鑫 林邦演 黄汉亿 李旻龙

(东莞市新一代人工智能产业技术研究院,广东 东莞 523867)

摘要:针对化工厂中化工袋种类繁多、遮挡干扰、放置复杂等因素,导致化工袋识别模型的定位效果较差、实时性不佳等问题,提出一种基于MBE-YOLOv5的轻量化化工袋目标检测方法。首先,用MobileNetV3网络替换YOLOv5的主干网络,降低模型的参数和运算量,提高模型的检测速度;然后,在YOLOv5的颈部网络引入双向特征金字塔网络结构进行多尺度特征融合,提高模型的识别准确率;最后,采用EIoU函数优化损失,提高模型的定位精度。实验结果表明,MBE-YOLOv5模型相比YOLOv5模型,参数量下降了37.7%,运算量降低了58.1%,检测速度提升了9.5%,mAP@0.5提高了0.7%;在检测速度和检测精度之间取得较好的平衡,能满足化工袋在线检测识别定位的要求。

关键词:YOLOv5模型;MobileNetV3网络;双向特征金字塔网络;EIoU函数;化工袋目标检测

中图分类号:TP391.41           文献标志码:A         文章编号:1674-2605(2024)02-0006-07

DOI:10.3969/j.issn.1674-2605.2024.02.006

Lightweight Chemical Bag Target Detection Method Based on MBE-YOLOv5 

LIU Weixin LIN Banyan HUANG Hanyi LI Minlong

(Dongguan Institute of New Generation Artificial Intelligence Industry Technology, Dongguan 523867, China)

Abstract: A lightweight chemical bag target detection method based on MBE-YOLOv5 is proposed to address the issues of poor positioning and real-time performance of chemical bag recognition models in chemical plants, which are caused by various types of chemical bags, occlusion interference, and complex placement. Firstly, replace the backbone network of YOLOv5 with MobileNetV3 network to reduce model parameters and computational complexity, and improve the detection speed of the model; Then, a bidirectional feature pyramid network structure is introduced into the neck network of YOLOv5 for multi-scale feature fusion to improve the recognition accuracy of the model; Finally, the EIoU function is used to optimize the loss and improve the positioning accuracy of the model. The experimental results show that compared to the YOLOv5 model, the MBE-YOLOv5 model reduces the number of parameters by 37.7%, the computational complexity by 58.1%, and the detection speed by 9.5%, mAP@0.5 Improved by 0.7%; Achieving a good balance between detection speed and accuracy can meet the requirements of online detection, recognition, and positioning of chemical bags.

Keywords: YOLOv5 model; MobileNetV3 network; bidirectional feature pyramid network; EIoU function; chemical bag target detection

打印