YOLOv8-DG:基于YOLOv8s改进的草莓成熟度检测模型  

YOLOv8-DG:An Improved Strawberry Ripeness Detection Model Based on YOLOv8s

作　　者：杨滨硕 狄巨星[1] 杨阳[1] YANG Binshuo;DI Juxing;YANG Yang(Hebei University of Architecture and Engineering,Zhangjiakou 075000,China)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《长江信息通信》2025年第1期82-86,共5页Changjiang Information & Communications

摘　　要：针对自然环境下的红熟期草莓,为实现其成熟度的高效检测,提出一种基于YOLOv8s改进的草莓成熟度检测模型:YOLOv8-DG。以YOLOv8s模型为基础,在C2f模块中引入DCNv2(Deformable Convolution v2)结构,提高了模型的鲁棒性和对特征的辨别性,并将损失函数替换为GIoU,提高模型收敛速度,从而提高模型的性能。实验结果表明,YOLOv8-DG模型GFLOPS仅有27.6,相比原YOLOv8s模型减少3%;且平均精确度较原YOLOv8s模型提高2.1个百分点。改进后模型相较当前主流YOLO系列模型,平均精度等指标均有所提升,基本可以满足自然环境下的草莓成熟度检测。To efficiently detect the maturity of red ripe strawberries in natural environments,an improved strawberry maturity detection model based on YOLOv8s,YOLOv8-DG,is proposed.Based on the YOLOv8s model,DCNv2(Deformable Convolution v2)structure is introduced into the C2f module to improve the robustness and feature discrimination of the model.The loss function is replaced with GIoU to improve the convergence speed of the model and thus enhance its performance.The experimental results showed that the GFLOPS of the YOLOv8-DG model was only 27.6,a decrease of 3%compared to the original YOLOv8s model;And the average accuracy is 2.1 percentage points higher than the original YOLOv8s model.Compared with the current mainstream YOLO series models,the improved model has improved in terms of average accuracy and other indicators,and can basically meet the requirements of strawberry maturity detection in natural environments.

关 键 词：YOLOv8s YOLOv8-DG DCNv2 模型收敛速度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]