基于机器视觉的白芍药材双面品质检测系统  

Double-sided Quality Detection System of Debark Peony RootBased on Machine Vision

作　　者：陈泽伟 乔印虎 周玉蝶 CHEN Zewei;QIAO Yinhu;ZHOU Yudie(College of Mechanical Engineering,Anhu Science and Technologyi University,Chuzhou 233100,China)

机构地区：[1]安徽科技学院机械工程学院,安徽滁州233100

出　　处：《常州工学院学报》2024年第2期1-7,共7页Journal of Changzhou Institute of Technology

基　　金：安徽省教育厅安徽高校自然科学研究项目(2022AH040238)。

摘　　要：为提高白芍品质检测的准确度和效率,提出一种基于机器视觉技术的双面白芍品质检测系统。通过镜面反射,系统对药材上下两面图像同时采集处理,以检测药材的品质。除对比分析单面和双面图像的识别效果外,还在YOLOv8s模型中引入Shuffle Attention注意力机制和Focal EIOU Loss损失函数优化算法。通过实验对比,白芍的单面识别平均精度均值为94.4%,双面识别的平均精度均值为92.8%,改进后的算法对双面白芍图像的识别平均精度均值达到99.3%,同时可避免样本不均衡引起的预测偏向性。实验结果验证了白芍药材双面品质检测系统的可行性和实用性。A double-sided quality inspection system for debark peony root based on machine vision technology is proposed to improve the accuracy and efficiency of quality testing.Utilizing the principle of mirror reflection,images of both sides of the medicinal materials are simultaneously captured and processed to detect their quality.By comparing the recognition performance of single-sided and dual-sided images,the YOLOv8s model introduces the Shuffle Attention mechanism and Focal EIOU Loss optimization algorithm.Experimental results show that the average recognition accuracy for single-sided debark peony root is 94.4%,while for the dual-sided,it is 92.8%.With the improved algorithm,the average recognition accuracy for dual-sided debark peony root images reaches 99.3%,while also mitigating prediction bias caused by sample imbalance.The experimental results verify the feasibility and practicality of the dual-sided quality inspection system for debark peony root medicinal materials.

关 键 词：机器视觉 双面检测 白芍药材 品质检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]