基于改进YOLOv11的水果成熟度检测
(1.太原师范学院 计算机科学与技术学院,山西 晋中 030619; 2.智能优化计算与区块链技术山西省重点实验室,山西 晋中 030619)
摘要
针对水果成熟度检测中存在的精度不足、复杂背景下识别难度大,传统方法在特征提取上表现出明显的局限性等问题,为此,提出了一种基于改进YOLOv11的水果成熟度检测算法(AGLU-YOLOv11),以满足水果成熟度检测中对数据高效、可靠采集的需求。AGLU-YOLOv11通过优化YOLOv11主干网络中的C3k2模块,融合CATM(Conv Additive Self-Attention)与CGLU(Convolutional Gated Linear Unit)设计了C3k2_AddBlock_CGLU模块,显著提升特征提取能力及多品种、多阶段成熟度果实的适应性。同时,在特征融合阶段引入AFCA注意力机制,强化全局特征表达及复杂背景的适应性,实现高效水果质量检测与标注。实验结果表明,AGLU-YOLOv11在Precision、Recall、mAP@0.5和mAP@0.5:0.95指标上相比其他检测模型,在精度、鲁棒性和多尺度目标适应性上表现更优,能够更好地满足识别水果成熟度的需求。
关键词
YOLO;目标检测;CGLU;CATM;水果成熟度检测
参考文献
- [1] 赵杰斌,邓浩,徐振林.基于食品安全指数法的台山市蔬菜和水果农药残留风险评估 [J].现代食品科技,2023,39(6):277-283.
- [2] 田有文,吴伟,卢时铅,等.深度学习在水果品质检测与分级分类中的应用 [J].食品科学,2021,42(19):260-270.
- [3] CUBERO S,ALEIXOS N,ALBERT F,et al. Optimised Computer Vision System for Automatic Pre-Grading of Citrus Fruit in the Field Using a Mobile Platform [J].Precision Agriculture,2014,15(1):80-94.
- [4] OHALI A Y. Computer Vision Based Date Fruit Grading System: Design and Implementation [J].Journal of King Saud University-Computer and Information Sciences,2010,23(1):29-36.
- [5] 宋立航,张屹,时寅豪.基于改进YOLOv5s的自然环境下油桃成熟度检测方法 [J].中国农机化学报,2024,45(9):250-257+2.
- [6] ZHAO J F,DU C F,LI Y,et al. YOLO-Granada: A Lightweight Attentioned Yolo for Pomegranates Fruit Detection [J/OL].Scientific Reports,2024,14(1):16848[2024-10-26].https://doi.org/10.1038/s41598-024-67526-4.
- [7] REDMON J. You Only Look Once, Unified,Real-Time Object Detection [C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas:IEEE,2016:779-788.
- [8] 常琳凯,马继东,张昊,等.基于改进YOLOv8的木材表面缺陷检测研究 [J/OL].林业工程学报,2024:1-12[2024-12-03].https://doi.org/10.13360/j.issn.2096-1359.202410002.
- [9] ARISTOTELES,SYARIF A,SUTYARSO,et al. Identification of Human Sperm Based on Morphology Using the You Only Look Once Version 4 Algorithm [J/OL].International Journal of Advanced Computer Science and Applications (IJACSA),2022,13(7)[2024-11-02].https://dx.doi.org/10.14569/IJACSA.2022.0130752.
- [10] ZHANG T,LI L,ZHOU Y,et al. CAS-ViT: Convolutional Additive Self-Attention Vision Transformers for Efficient Mobile Applications [J/OL].arXiv:2408.03703 [cs.CV].[2024-10-28].https://arxiv.org/abs/2408.03703?context=cs.
- [11] SHI D. TransNeXt: Robust Foveal Visual Perception for Vision Transformers [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle:IEEE,2024:17773-17783.
- [12] HU J,SHEN L,SUN G. Squeeze-and-Excitation Networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2019,42(8):2011-2023.
- [13] SUN H,WEN Y,FENG H,et al. Unsupervised Bidirectional Contrastive Reconstruction and Adaptive Fine-Grained Channel Attention Networks for Image Dehazing [J/OL].Neural Networks,2024,176:106314[2024-11-02].https://doi.org/10.1016/j.neunet.2024.106314.
基金项目
山西省科技战略研究专项重点项目(202304031401011)
作者简介
赵鹏(1973—),男,汉族,山西太原人,教授,博士在读,研究方向:软件工程、大数据、区块链;强光磊(2001—),男,汉族,安徽芜湖人,硕士研究生在读,研究方向:区块链、目标检测;通信作者:卢波(1988—),男,汉族,山西晋中人,博士,研究方向:区块链、协议安全性分析;高扬(2000—),男,汉族,山西忻州人,硕士研究生在读,研究方向:图像处理、目标检测;张仟祥(2000—),男,汉族,山西太原人,硕士研究生在读,研究方向:特征识别、嵌入式与物联网工程。