高级检索

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于全卷积神经网络的植物叶片分割算法

胡静 陈志泊 杨猛 张荣国 崔亚稷

胡静, 陈志泊, 杨猛, 张荣国, 崔亚稷. 基于全卷积神经网络的植物叶片分割算法[J]. 北京林业大学学报, 2018, 40(11): 131-136. doi: 10.13332/j.1000-1522.20180007
引用本文: 胡静, 陈志泊, 杨猛, 张荣国, 崔亚稷. 基于全卷积神经网络的植物叶片分割算法[J]. 北京林业大学学报, 2018, 40(11): 131-136. doi: 10.13332/j.1000-1522.20180007
Hu Jing, Chen Zhibo, Yang Meng, Zhang Rongguo, Cui Yaji. Plant leaf segmentation method based on fully convolutional neural network[J]. Journal of Beijing Forestry University, 2018, 40(11): 131-136. doi: 10.13332/j.1000-1522.20180007
Citation: Hu Jing, Chen Zhibo, Yang Meng, Zhang Rongguo, Cui Yaji. Plant leaf segmentation method based on fully convolutional neural network[J]. Journal of Beijing Forestry University, 2018, 40(11): 131-136. doi: 10.13332/j.1000-1522.20180007

基于全卷积神经网络的植物叶片分割算法

doi: 10.13332/j.1000-1522.20180007
基金项目: 

国家自然科学基金项目 61402038

详细信息
    作者简介:

    胡静,博士生,副教授。主要研究方向:图像处理与模式识别。Email:279641292@qq.com 地址:030006山西省太原市万柏林区窊流路66号太原科技大学计算机学院

    责任作者:

    陈志泊,教授,博士生导师。主要研究方向:数据库技术、计算机软件与理论。Email:zhibo@bjfu.edu.cn 地址:100083北京市海淀区清华东路35号北京林业大学信息学院

  • 中图分类号: TP391.41

Plant leaf segmentation method based on fully convolutional neural network

  • 摘要: 目的植物叶片分割旨在从背景中分割出叶片区域,去除背景对象干扰。这对植物病害识别和物种鉴定具有重大意义。方法本文设计了基于全卷积神经网络的植物叶片分割算法。首先,目标函数用对数逻辑函数代替复杂的Softmax多类预测函数,从而将分割任务转化为适合于植物叶片分割的二分类问题;其次,把批归一化技术引入全卷积神经网络,从而改善网络整体的收敛性。最后,针对当前植物叶片分割研究中缺乏评估指标的状况,设计了新的评估协议——受试者工作特征曲线,该曲线反映了不同阈值情况下植物叶片图像分割的召回率与误报率之间的变化情况。结果本文提出的算法降低了全卷积神经网络的参数复杂度,改善了网络的收敛性。实验结果表明,该方法比Leafsnap提到的基于颜色的分割方法更完整地分割了植物叶片区域;提出的ROC曲线能够充分评估植物叶片的分割性能。结论与传统方法相比,基于深度学习的植物叶片分割方法实现了输入图像的端对端处理,无需图像转换、噪声滤波和形态运算等预处理技术,因此在植物叶片分割上具有可行性。

     

  • 图  1  分割前后的植物叶片图像

    Figure  1.  Plant leaf images before and after segmentation

    图  2  基于全卷积神经网络植物叶片图像分割方法框架

    Figure  2.  Framework of plant leaf image segmentation method based on fully convolutional neural network

    图  3  来自Leafsnap数据库的所有184种树种的缩略图

    此图源自文献[3] This image is derived from reference [3].

    Figure  3.  Thumbnail of all 184 tree species from the Leafsnap database

    图  4  基于全卷积神经网络的植物叶分割方法与基于颜色的分割方法的比较

    Figure  4.  Comparison between plant leaf segmentation method based on fully convolutional neural network and that based on the color segmentation method

    图  5  基于全卷积神经网络的植物叶片图像分割方法所产生的ROC曲线

    Figure  5.  ROC curve generated by plant leaf segmentation method based on fully convolutional neural network

    表  1  本文提出的卷积神经网络的参数配置

    Table  1.   Parameter configuration of the proposed fully convolutional neural network

    CBM学习块
    CBM learning block
    名称
    Name
    激活函数
    Activiate function
    滤波器(h×w×c×g)
    Filter (h×w×c×g)
    步进
    Step
    池化操作
    Pooling operation
    卷积层1 Convolutional layer 1 (conv 1) 3×3×3×64 1
    模块1 Block 1 批归一化层1 Batch normalization layer 1 ReLU 1
    最大池化层1 Max. pooling layer 1 2 3×3
    卷积层2 Convolutional layer 2 (conv 2) 3×3×64×128 1
    模块2 Block 2 批归一化层2 Batch normalization layer 2 ReLU 1
    最大池化层2 Max. pooling layer 2 2 3×3
    卷积层3 Convolutional layer 3(conv 3) 3×3×128×256 1
    模块3 Block 3 批归一化层3 Batch normalization layer 3 ReLU 1
    最大池化层3 Max. pooling layer 3 2 3×3
    卷积层4 Convolutional layer 4 (conv 4) 3×3×256×512 1
    模块4 Block 4 批归一化层4 Batch normalization layer 4 ReLU 1
    最大池化层4 Max. pooling layer 4 2 3×3
    卷积层5 Convolutional layer 5 (conv 5) 3×3×512×512 1
    模块5 Block 5 批归一化层5 Batch normalization layer 5 ReLU 1
    最大池化层5 Max. pooling layer 5 2 3×3
    卷积层6 Convolutional layer 6 (conv 6) 1×1×512×1 024
    注:hwcg分别表示高度、宽度、通道和组的大小。Notes:hwcg indicate height, width, channel and group size, respectively.
    下载: 导出CSV

    表  2  阈值为0时误报率对应的召回率

    Table  2.   Recall rate corresponding to false alarm rate when threshold is 0

    误报率False alarm rate 0.1% 0.67% 1% 10%
    召回率Recall rate 80.45% 95.80% 96.62% 99.68%
    下载: 导出CSV
  • [1] Cope J S, Corney D, Clark J Y, et al.Plant species identification using digital morphometrics:a review[J].Expert Systems with Applications, 2012, 39(8):7562-7573. doi: 10.1016/j.eswa.2012.01.073
    [2] Clarke J, Barman S, Remagnino P, et al.Venation pattern analysis of leaf images[C]//Bebis G, Boyle R, Parvin B, et al.Advances in visual computing: part Ⅱ. Berlin: Springer, 2006: 427-436.
    [3] Kumar N, Belhumeur P N, Biswas A, et al.Leafsnap: a computer vision system for automatic plant species identification[C]//Fitzgibbon A, Lazebnik S, Perona P, et al. Computer vision: ECCV 2012. Berlin: Springer, 2012: 502-516.
    [4] Wang X F, Huang D S, Du J X, et al.Classification of plant leaf images with complicated background[J].Applied Mathematics & Computation, 2008, 205(2):916-926. http://cs231n.stanford.edu/reports/2017/pdfs/325.pdf
    [5] Lee S H, Chan C S, Wilkin P, et al.Deep-plant: plant identification with convolutional neural networks[C]//Proccedings of IEEE International Conference on Image Processing.Piscataway: IEEE Signal Processing Society, 2015: 452-456.
    [6] Kavitha P, Ananthi B.Segmentation of unhealthy region of plant leaf using image processing techniques:a survey[J].International Journal of Research in Engineering and Technology, 2014, 11(3):24-27. https://ieeexplore.ieee.org/document/7164858/
    [7] Boykov Y, Funka-Lea G. Graph cuts and efficient N-D image segmentation[J].International Journal of Computer Vision, 2006, 70(2):109-131. doi: 10.1007/s11263-006-7934-5
    [8] Shi J, Malik J. Normalized cuts and image segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8): 888-905. doi: 10.1109/34.868688
    [9] Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm[J].IEEE Transactions on Medical Imaging, 2001, 20(1):45-57. doi: 10.1109/42.906424
    [10] Valliammal N, Geethalakshmi S N. A novel approach for plant leaf image segmentation using fuzzy clustering[J].International Journal of Computer Applications, 2012, 44(13):10-20. doi: 10.5120/6322-8669
    [11] Valliammal N, Geethalakshmi S N. Leaf image segmentation based on the combination of wavelet transform and k means clustering[J].International Journal of Advanced Research in Artificial Intelligence, 2012, 1(3):25-39. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=Doaj000003577404
    [12] Shelhamer E, Long J, Darrell T, et al.Fully convolutional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(4):640-651. doi: 10.1109/TPAMI.2016.2572683
    [13] 张帅, 淮永建.基于分层卷积深度学习系统的植物叶片识别研究[J].北京林业大学学报, 2016, 38(9):108-115. http://bjly.chinajournal.net.cn/WKC/WebPublication/paperDigest.aspx?paperID=94fda7a0-37de-449e-bcb0-bfb6a7916e6c

    Zhang S, Huai Y J.Leaf image recognition based on layered convolutions neural network deep learning[J].Journal of Beijing Forestry University, 2016, 38(9):108-115. http://bjly.chinajournal.net.cn/WKC/WebPublication/paperDigest.aspx?paperID=94fda7a0-37de-449e-bcb0-bfb6a7916e6c
    [14] Noh H, Hong S, Han B.Learning deconvolution network for semantic segmentation[C/OL]//Proceedings of IEEE International Conference on Computer Vision. Washington D C: IEEE Computer Society, 2015[2017-10-10]. https://arxiv.org/pdf/1505.04366.pdf.
    [15] Chen L C, Papandreou G, Kokkinos I, et al.Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4):834-848. doi: 10.1109/TPAMI.2017.2699184
    [16] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[C]//Pereira F, Burges C J C, Bottou L, et al. Advances in neural information processing systems. New York: Curran Associates Inc., 2012: 1097-1105.
    [17] Ioffe S, Szegedy C.Batch normalization: accelerating deep network training by reducing internal covariate shift[C/OL]. ICML'15 Proccedings of the 32nd International Conference on Machine Learning, 2015, 37: 448-456[2017-09-20]. https://arxiv.org/pdf/1502.03167.pdf.
  • 加载中
图(5) / 表(2)
计量
  • 文章访问数:  3183
  • HTML全文浏览量:  332
  • PDF下载量:  34
  • 被引次数: 0
出版历程
  • 收稿日期:  2018-01-04
  • 修回日期:  2018-09-10
  • 刊出日期:  2018-11-01

目录

    /

    返回文章
    返回