基于改进YOLOV8n的食用玫瑰花轻量化识别方法

打开文本图片集

中图分类号：TP242 文献标识码：A 文章编号：2095-2953（2025）12-0012-06

Lightweight Recognition Method for Edible Roses Based on Improved YOLOV8n

YANG Lu ，LI Kun-zhen，YANG Yun- fu ，CHEN Wen-hong， WANG Feng （1.ColegeofMechanicalEngineringandTransportation，SouthwestForestry University，Kunming Yunnan5O224，Chna; 2.Yunnan Agricultural Machinery Research Institute Co.，Ltd.，Kunming Yunnan 65O224，China）

Abstract：In order to improve the recognition accuracy of edible roses incomplex field environment and realize the lightweight model，this study proposesan improved method basedonthe YOLOv8ndetectionmodel.Firstly，inorder to efectively reduce the number of parameters and model volume，MobileNetV3 lightweight network is used to replace the backbone feature extraction network of the original model to extract feature information，which facilitates detection in complex field environment.Secondly，GSConv is integrated with C2f module in the neck network to reduceredundant computations while retaining strong feature expresion capability Finally，the loss functionof YOLOv8n is replaced with PIoU to solve the problem of bounding box expansion and slow convergence speed，enabling themodel toachieve faster convergence speed.The experimental results show thatthe improved lightweight model has 2.49 M parameters，with a detection accuracy 20.9% higher than that of the original YOLOv8n model. The size of the model is4.75MB，which is1.19MB lowerthanthatof theoriginalmodel.The precisionandrecalrate are 96.6% and 95.40% respectively. This model improves the detection accuracy of edible roses while achieving lightweight，providing a theoretical basis for the subsequent intellgent and efficient picking of edible roses.

Key words： ΥOLOv8n ；edible roses；mobilenetv3；lightweight

玫瑰花为蔷薇科植物，除具有食用价值外还具有药用价值。（剩余6919字）