基于EfficientDet-D1的草莓快速检测及分类
CSTR:
作者:
作者单位:

1.仲恺农业工程学院自动化学院,广州510225;2.广东省农产品冷链运输与物流工程技术研究中心, 广州510225;3.仲恺农业工程学院机电工程学院,广州510225;4.华中农业大学工学院/国家柑橘保鲜技术研发专业中心,武汉430070;5.粤港澳大湾区农产品数字物流研究中心, 广州510225

作者简介:

张小花, E-mail:janemmy2000@163.com

通讯作者:

李善军, E-mail:shanjunlee@mail.hzau.edu.cn

中图分类号:

S668.4;TP391.41

基金项目:

国家重点研发计划项目(2020YFD1000101);广东省普通高校特色创新类项目(2019KTSCX064);广州市科技计划项目(202002020028);广州市科信局项目GZKTP202003);广东省农产品保鲜物流共性关键技术研发创新团队(2021KJ145


Rapid detection and classification of strawberries based on EfficientDet-D1
Author:
Affiliation:

1.College of Automation,Zhongkai University of Agriculture and Engineering, Guangzhou 510225,China;2.Guangdong Agricultural Products Cold Chain Transportation and Logistics Engineering Technology Research Center,Guangzhou 510225,China;3.College of Mechanical Engineering,Zhongkai University of Agriculture and Engineering,Guangzhou 510225,China;4.College of Engineering/National R&D Center for Citrus Preservation, Huazhong Agricultural University,Wuhan 430070,China;5.Guangdong-Hong Kong-Macao Greater Bay Area Agricultural Products Digital Logistics Research Center,Guangzhou 510225,China

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献 [17]
  • | | | |
  • 文章评论
    摘要:

    为了快速识别自然环境下的成熟草莓与未成熟草莓,本研究提出了基于EfficientDet-D1的草莓快速检测及分类方法。该方法具有EfficientNet 网络中快速归一化特征加权融合特点,应用该方法与YOLOv3、YOLOv4、Faster-RCNN以及EfficientDet-D0模型进行对比试验,结果显示,YOLOv3、YOLOv4、Faster-RCNN、EfficientDet-D0和EfficientDet-D1等5种算法的平均精度均值(PmA)分别为 89.51%、69.02%、96.54%、96.71%、97.50%。试验结果表明,EfficientDet-D1在成熟草莓与未成熟草莓的检测性能均优于其他4种目标检测算法,有较好的泛化性和鲁棒性,且使用模型参数量较小的EfficientNet网络,更适合作用于移动端识别,可实现草莓快速识别中的速度与精度要求。

    Abstract:

    A target detection algorithm based EfficientDet-D1 model was proposed to meet the requirements of speed and accuracy in the rapid detection and classification of strawberries using the fast normalized feature weighted fusion feature in EfficientNet network to quickly identify the ripe and unripe strawberries in natural environments.The YOLOv3,YOLOv4,Faster-RCNN and EfficientDet-D0 models were used for comparative experiments.The results showed that mean average precision (mAP) of the five algorithms including YOLOv3,YOLOv4,Faster-RCNN,EfficientDet-D0 and EfficientDet-D1 was 89.51%,69.02%,96.54%,96.71%,and 97.50%,respectively.The detection performance of EfficientDet-D1 in the ripe and unripe strawberries is better than that of the other four target detection algorithms,which has better generalization and robustness,By using the EfficientNet network with a small number of model parameters.It is more suitable for mobile identification,and can provide a new solution for the automatic picking technology of the ripe strawberries.

    表 3 未成熟草莓检测的不同算法各项评价指标对比Table 3 Comparison of evaluation indicators of different algorithms for the detection of unripe strawberries
    表 1 草莓样本数分布情况Table 1 Distribution of strawberry samples
    表 2 成熟草莓检测的不同算法各项评价指标对比Table 2 Comparison of evaluation indicators of different algorithms for the detection of ripe strawberries
    图1 数据增强方法Fig.1 Data augmentation method
    图2 EfficientDet主干特征提取网络Fig.2 EfficientDet backbone feature extraction network
    图3 加强特征提取网络结构Fig.3 Strengthen feature extraction network structure
    图4 试验整体流程Fig.4 The overall flow of experiment
    图5 5种检测算法效果图Fig.5 Effect diagram of five detection algorithms
    参考文献
    [1] 贾宗维,姚思敏,张如意,等.基于改进U-Net网络的大棚草莓果实识别[J].山西农业大学学报(自然科学版),2022,42(2):120-128.JIA Z W,YAO S M,ZHANG R Y,et al. Recognition of strawberry fruits in greenhouses based on improved U-Net network [J].Journal of Shanxi Agricultural University (natural science edition),2022,42(2):120-128(in Chinese with English abstract).
    [2] 刘小刚,范诚,李加念,等.基于卷积神经网络的草莓识别方法[J].农业机械学报,2020,51(02):237-244.LIU X G,FAN C,LI J N,et al.Strawberry recognition method based on convolutional neural network [J].Journal of the CSAM,2020,51(2):237-244(in Chinese with English abstract).
    [3] 许丽建.颜色差值分类算法在草莓成熟程度检测中的应用[J].科技通报,2012,28(10):160-162.XU L J.Color difference in strawberry classification algorithm of maturity of the application of the test[J].Bulletin of science and technology,2012,28(10):160-162(in Chinese with English abstract).
    [4] 覃磊,孙开琼,李诗高,等.基于RGB颜色相似度的成熟草莓图像分割[J].浙江农业学报,2016,28(2):330-337.QIN L,SUN K Q,LI S G,et al.Image segmentation of ripe strawberry based on RGB color similarity[J].Acta agriculturae zhejiangensis,2016,28(2):330-337(in Chinese with English abstract).
    [5] 蒋浩,张初,刘飞,等.基于高光谱图像多光谱参数的草莓成熟度识别[J].光谱学与光谱分析,2016,36(5):1423-1427.JIANG H,ZHANG C,LIU F,et al.Identification of strawberry ripeness based on multispectral indexes extracted from hyperspectral images[J].Spectroscopy and spectral analysis,2016,36(5):1423-1427(in Chinese with English abstract).
    [6] 王雪光,陈淑红.基于K均值聚类的成熟草莓图像分割算法[J].农机化研究,2013,35(6):51-54.WANG X G,CHEN S H.The ripe strawberry image segmentation algorithm based on K-means clustering[J].Journal of agricultural mechanization research,2013,35(6):51-54(in Chinese with English abstract).
    [7] 赵玲,周桂红.基于颜色特征的草莓成熟度识别技术研究[J].河北农业大学学报,2017,40(2):97-101.ZHAO L,ZHOU G H.Study on recognition technology of strawberry maturity based on color feature[J].Journal of Hebei Agricultural University,2017,40(2):97-101(in Chinese with English abstract).
    [8] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Communications of the acm,2017,60(6):84-90.
    [9] 杨万里,段凌凤,杨万能.基于深度学习的水稻表型特征提取和穗质量预测研究[J].华中农业大学学报,2021,40(1):227-235.YANG W L,DUAN L F,YANG W N.Deep learning-based extraction of rice phenotypic characteristics and prediction of rice panicle weight[J].Journal of Huazhong Agricultural University,2021,40(1):227-235(in Chinese with English abstract).
    [10] ZHOU X,LEE W S,AMPATZIDIS Y,et al.Strawberry maturity classification from UAV and near-ground imaging using deep learning[J/OL].Smart agricultural technology,2021,1:100001[2022-04-13].https://doi.org/10.1016/j.atech.2021.100001.
    [11] TAN M X,PANG R M,LE Q V.EfficientDet:scalable and efficient object detection[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Seattle,WA,USA.Seattle :IEEE,2020:10778-10787.
    [12] REN S Q,HE K M,GIRSHICK R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE transactions on pattern analysis and machine intelligence,2017,39(6):1137-1149.
    [13] SHINDE S,KOTHARI A,GUPTA V.YOLO based human action recognition and localization[J].Procedia computer science,2018,133:831-838.
    [14] 王季.基于EfficientDet的全球小麦麦穗检测方法研究[J].农业装备与车辆工程,2021,59(12):108-111.WANG J.Research on global wheat head detection method based on EfficientDet[J].Agricultural equipment & vehicle engineering,2021,59(12):108-111(in Chinese with English abstract).
    [15] WU D H,LV S C,JIANG M,et al.Using channel pruning-based YOLO v4 deep learning algorithm for the real-time and accurate detection of apple flowers in natural environments[J/OL].Computers and electronics in agriculture,2020,178:105742[2022-04-13].https://doi.org/10.1016/j.compag.2020.105742.
    [16] LIN T Y,DOLLáR P,GIRSHICK R,et al.Feature pyramid networks for object detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition.Honolulu,HI,USA.Honolulu:IEEE,2017:936-944.
    [17] LIU S,QI L,QIN H F,et al.Path aggregation network for instance segmentation[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City,UT,USA.Salt Lake City :IEEE,2018:8759-8768.
    相似文献
    引证文献
    网友评论
    网友评论
    分享到微博
    发 布
引用本文

张小花,李浩林,李善军,张文峰,冼镇鸿.基于EfficientDet-D1的草莓快速检测及分类[J].华中农业大学学报,2022,41(6):262-269

复制
分享
文章指标
  • 点击次数:1087
  • 下载次数: 1206
  • HTML阅读次数: 239
  • 引用次数: 0
历史
  • 收稿日期:2022-04-13
  • 在线发布日期: 2022-12-09
文章二维码