改进YOLOv8的果园葡萄检测算法
Improved orchard grape detection algorithm of YOLOv8
王成健 1徐振平 1文汉云1
作者信息
- 1. 长江大学计算机科学学院,荆州 434000
- 折叠
摘要
针对现有的果园葡萄检测算法在针对枝干遮挡、果实遮挡时检测精度不高的问题,提出一种改进YOLOv8n的果园葡萄检测算法.首先,向主干网络末端和颈部网络引入CA注意力机制,以提高模型对关键特征的关注度;然后使用MPDIoU替换原模型中的CIoU损失函数,加快模型收敛速度并提高边界框回归精度.实验结果表明,改进后的模型在P、R和mAP上均有提升,能够满足复杂果园环境下对葡萄检测的需求.
Abstract
Due to the low detection accuracy of existing orchard grape detection algorithms in the case of stem occlusion and fruit occlusion,an improved YOLOv8n orchard grape detection algorithm was proposed.Firstly,CA attention mechanism is intro-duced to the end of backbone network and neck network to improve the model's attention to key features.Then MPDIoU was used to replace the CIoU loss function in the original model to accelerate the convergence speed and improve the precision of bounding box regression.The experimental results show that the improved model has improved in P,R and mAP,and can meet the needs of grape detection in complex orchard environment.
关键词
葡萄检测/YOLOv8n/注意力机制/损失函数/MPDIoUKey words
grape detection/YOLOv8n/attention mechanism/loss fuction/MPDIoU引用本文复制引用
基金项目
国家自然科学基金资助项目(41372155)
出版年
2024
NETL
NSTL
万方数据