• 首页
  • 期刊简介
  • 编委会
  • 投稿指南
  • 收录情况
  • 杂志订阅
  • 联系我们
引用本文:郑 昊,魏霖静.基于深度学习的咖啡果实成熟度检测方法[J].软件工程,2025,28(2):32-37.【点击复制】
【打印本页】   【下载PDF全文】   【查看/发表评论】  【下载PDF阅读器】  
←前一篇|后一篇→ 过刊浏览
分享到: 微信 更多
基于深度学习的咖啡果实成熟度检测方法
郑 昊,魏霖静
(甘肃农业大学信息科学技术学院,甘肃 兰州 730070)
vzhenghao@163.com; 916277964@qq.com
摘 要: 针对目前咖啡果实的成熟度主要依赖人工判断且果实相互遮挡导致识别难度大的问题,文章提出一种改进的YOLOv8模型的咖啡果实成熟度检测方法。首先,在Backbone端使用iRMB(InvertedResidualMobile Block)混合网络模块替换C2f(CSPBottleneckwith2Convolutions),增强模型特征表示能力;其次,引入BiFormer 注意力机制,增强对遮挡和小目标果实的检测能力,更换CARAFE(Content-AwareReAssemblyofFeatures)上采样算子,拓宽感受视野;最后,引入Wise-IOU损失函数,加速模型收敛。经验证,相比于原算法,改进算法的精确率、召回率、平均精确率分别提升了6.1百分点、1.0百分点、2.9百分点。研究结果表明,改进的YOLOv8模型可以为咖啡果实的成熟度检测提供有效参考。
关键词: 咖啡果实;成熟度检测;YOLOv8;CARAFE;BiFormer
中图分类号: TP391    文献标识码: A
基金项目: 2021年度兰州市人才创新创业项目(2021-RC-47);2022年度科技部国家外专项目(G2022042005L);2023年甘肃省高等学校产业支撑项目(2023CYZC-54);2023年甘肃省重点研发计划(23YFWA0013);2023年甘肃农业大学美育和劳动教育教学改革项目(2023-09)
A Deep Learning-Based Method for Detecting the Ripeness of Coffee Fruits
ZHENG Hao, WEI Linjing
(College of Information Science and Technology, Gansu Agricultural University, Lanzhou 730070, China)
vzhenghao@163.com; 916277964@qq.com
Abstract: This paper addresses the current challenges in determining the ripeness of coffee fruits, which primarily relies on manual judgement and is complicated by the occlusion of fruits. To meet these challenges, an improved YOLOv8 model for detecting the ripeness of coffee fruits is proposed. Firstly, iRMB (Inverted Residual Mobile Block) mixed network module is used to replace the C2f (CSP Bottleneck with 2 Convolutions), enhancing the model's feature representation ability. Secondly, BiFormer attention mechanism is introduced to improve the detection capability for occluded and small coffee fruits, and the CARAFE (Content-Aware ReAssembly of Features) upsampling operator is replaced to widen the receptive field. Finally, Wise-IoU loss function is incorporated to accelerate model convergence. Validation results indicate that compared to the original algorithm, the improved algorithm shows increases of 6.1 percentage points in precision, 1.0 percentage points in recall, and 2.9 percentage points in mAP (mean Average Precision), respectively. Research findings suggest that the improved YOLOv8 model provides an effective reference for detecting the ripeness of coffee fruits.
Keywords: coffee fruit; ripeness detection; YOLOv8; CARAFE; BiFormer


版权所有:软件工程杂志社
地址:辽宁省沈阳市浑南区新秀街2号 邮政编码:110179
电话:0411-84767887 传真:0411-84835089 Email:semagazine@neusoft.edu.cn
备案号:辽ICP备17007376号-1
技术支持:北京勤云科技发展有限公司

用微信扫一扫

用微信扫一扫