基于YOLOv8的速食粉面多品种外包装识别方法  

作　　者：梁国健 周树文 黎东健 何冬康 

机构地区：[1]柳州职业技术大学,广西柳州545006

出　　处：《科技创新与应用》2025年第10期142-145,共4页Technology Innovation and Application

基　　金：2022年广西高校中青年教师科研基础能力提升项目(22023KY1085);2023年度广西职业教育教学改革研究项目(GXGZJG2023B209)。

摘　　要：该研究旨在解决速食粉面生产过程中对多品种外包装的智能识别难题,计算机视觉技术为速食粉面外包装生产线提供了一种高效、智能化的包装识别与分拣解决方案。当前,生产线上的包装识别系统往往受限于识别种类,每增加一种外包装识别功能,便大幅增加企业的生产开发成本。该文创新性地融合YOLOv8目标识别算法与计算机视觉技术,实现了速食粉面外包装的实时识别功能,并对外包装内食品质量无实质性影响,不仅减少了因扩展识别种类而增加的成本,还显著提升了生产线的运行效率。经实验表明,模型的精准度为99.93%,召回率为99.98%,mAP0.5为99.49%,mAP0.5∶0.95为99.47%,该研究给速食粉面包装生产企业提供了一种多品种外包装识别解决方案,具有一定的研究价值。This study aims to address the challenge of intelligent recognition of multi-variety packaging in the production process of instant noodles and noodles.Computer vision technology provides an efficient and automated packaging recognition and sorting solution for the production line of instant noodles and noodles packaging.Currently,packaging recognition systems on production lines are often limited in the types of recognition,and each additional packaging recognition function significantly increases the production and development costs for enterprises.This paper innovatively integrates the YOLOv8 object recognition algorithm with computer vision technology to achieve real-time recognition of instant noodles and noodles packaging,without substantially affecting the quality of the food inside the packaging.This not only reduces the increased cost caused by expanding the identification types,but also significantly improves the operating efficiency of the production line.Experiments show that the accuracy of the model is 99.93%,the recall rate is 99.98%,mAP0.5 of 99.49%,and mAP0.5∶0.95 of 99.47%.This research provides instant noodle packaging manufacturers with a multi-variety outer packaging identification solution,which has certain research value.

关 键 词：速食粉面 YOLOv8 计算机视觉 目标识别 深度学习 

分 类 号：TS213.2[轻工技术与工程—粮食、油脂及植物蛋白工程]