摘要
本文提出一种基于改进YOLOv8的检测分拣模型,提高芡果分拣的效率和准确性。以芡种为对象,采集带衣皮的芡种图像并进行预处理,通过在YOLOv8基础模型上引入SE模块提升特征提取能力,再通过使用WIoU函数来进一步提高模型精度,并进行多种模型对比试验。改进后的模型平均精确率达到99.6%,召回率达到99.8%,均值平均精度达到99.4%,比YOLOv8原始模型的精度分别提高了1.8%、1.6%、1.3%。提出基于改进YOLOv8的芡种成熟度检测模型在保持轻量化的同时,有较高的检测精度,可实现对不同成熟度的芡种进行识别并准确分类。
An improved detection and sorting model based on YOLOv8 was proposed,improving the accuracy and efficiency in maturity sorting of Euryale ferox.Taking the Euryale ferox seeds as the object,preprocessing was made by the collected images of Euryale ferox seeds with coat skin,the SE module was introduced into the basic YOLOv8 model to improve the feature extraction capability,and WIoU function was used to further improve the model accuracy,and multiple model comparison tests were conducted.After the improvement,the average precision(P),recall(R)and mean average precision(mAP)reached 99.6%,99.8%,99.4%,respectively.Compared with the original YOLOv8 model,the accuracy was improved by 1.8%,1.6%,1.3%,respectively.A maturity detection model based on the improved YOLOv8 could identify and accurately classify the varieties of different maturity while maintaining light weight and high detection accuracy.
作者
陈龙梅
张春雨
CHEN Longmei;ZHANG Chunyu(College of Mechanical Engineering,Anhui Science and Technology University,Fengyang 233100,China)
出处
《安徽科技学院学报》
2025年第1期70-76,共7页
Journal of Anhui Science and Technology University
基金
安徽省高校协同创新项目(GXXT-2019-036,GXXT-2022-077)。
关键词
YOLOv8
目标检测
注意力机制
深度学习
芡种
YOLOv8
Target detection
Attention mechanism
Deep learning
Euryale ferox seeds
