[1]刘嘉政,王雪峰*,王 甜.基于深度学习的树种图像自动识别[J].南京林业大学学报(自然科学版),2020,44(01):138-143.[doi:10.3969/j.issn.1000-2006.201809004]
 LIU Jiazheng,WANG Xuefeng*,WANG Tian.Automatic identification of tree species based on deep learning[J].Journal of Nanjing Forestry University(Natural Science Edition),2020,44(01):138-143.[doi:10.3969/j.issn.1000-2006.201809004]
点击复制

基于深度学习的树种图像自动识别
分享到:

《南京林业大学学报(自然科学版)》[ISSN:1000-2006/CN:32-1161/S]

卷:
44
期数:
2020年01期
页码:
138-143
栏目:
研究论文
出版日期:
2020-01-15

文章信息/Info

Title:
Automatic identification of tree species based on deep learning
文章编号:
1000-2006(2020)01-0138-07
作者:
刘嘉政王雪峰*王 甜
(中国林业科学研究院资源信息研究所,北京 100091)
Author(s):
LIU Jiazheng WANG Xuefeng* WANG Tian
(Research Institute of Resource Information Techniques, Chinese Academy of Forestry, Beijing 100091, China)
关键词:
深度学习方法 树种图像 卷积神经网络 自动识别
Keywords:
deep learning method tree species image convolution neural network automatic identification
分类号:
TP391.41
DOI:
10.3969/j.issn.1000-2006.201809004
文献标志码:
A
摘要:
【目的】为了探究深度学习方法用于林业树种图像智能识别的可行性,提出一种基于深度学习方法的自动识别树种新方法。在TensorFlow框架下,对卷积神经网络(CNN)模型进行改进,对7类树种图像进行自动识别研究。【方法】首先,在图像库建立时,为增加特征选择多样性,选择树木的树皮和树叶图像,保留自然背景; 另外,考虑到同一树种在不同树龄条件下树皮图像存在差异,因此加入不同树龄的树皮图像,并用胸径指标来表示树龄大小。其次,对每类树种图像随机挑选100张作为测试集,剩余数据集全部作为训练集。通过反复试验比较不同CNN结构设置、卷积层数量、全连接层层数、学习率等对结果的影响。采用Adam算法代替传统的随机梯度下降(SGD)算法,对模型进行优化,用指数衰减法对学习率进行调节,在交叉熵函数中加入L2正则项对权重进行惩罚,并采用Dropout策略和ReLU激励函数,以避免训练过程中过拟合现象。最后,确定适合试验要求的13层CNN结构,同时比较深度学习方法和传统人工特征识别方法的差异,与已有的树种图像识别方法做对比。【结果】提出的13层树种图像识别模型,对训练集和测试集取得了理想的识别效果,识别率分别为96.78%、91.89%,在未参与训练的验证集上取得了96%的平均准确率。相对于已有的人工特征识别方法,所提出的方法识别效率和准确度更高。【结论】基于改进的卷积神经网络树种识别模型识别效果明显高于传统方法,说明所提出的方法能够应用于树种识别,可为林业树种图像自动识别提供一条新思路。
Abstract:
【Objective】 In order to examine the feasibility of developing a deep learning method for the intelligent identification of forestry tree species, we propose a new method of automatic tree recognition tree based on deep learning. In this study, we improved a convolutional neural network model under the framework of TensorFlow, and used this to examine the automatic identification of images of seven tree species. 【Method】Initially, when creating an image library, in order to increase the feature selection diversity, the bark and leaf images of the trees are selected to preserve the natural background. In addition, considering that the same tree species can have different bark images depending on tree age, the images of bark from trees of different ages are added, and the DBH index is used to indicate the age of the tree. Thereafter, 100 images of each tree species were randomly selected as test sets, and the remaining data sets were all used as training sets. Through repeated experiments to compare the effects of different convolutional neural network structure settings, the number of convolutional layers, the number of fully connected layers, and the learning rate on the experimental results, we propose the use the Adam algorithm instead of the traditional SGD algorithm to optimize the model and use the index. The attenuation method adjusts the learning rate. The L2 regular term is added to the cross entropy function to penalize the weight, and the Dropout strategy and ReLU excitation function are used. Finally, a 13-layer convolutional neural network structure suitable for the experimental requirements of this study was determined. In addition, in order to compare differences between the deep learning method and the traditional artificial feature recognition method, we compared performance with that of an existing tree species image recognition method. 【Result】 The 13-layer tree species image recognition model proposed in this study achieved an ideal recognition effect on both training and test sets, with recognition rates of 96.78% and 91.89%, respectively. Furthermore, we obtained an average 96% recognition for the verification set that was not included in the training, and overall found the recognition efficiency and accuracy to be higher than those of the existing artificial feature recognition methods. 【Conclusion] Compared with the traditional method, the proposed method, based on an improved convolutional neural network tree species recognition model, can be applied to tree species identification and provides a new approach for forestry tree species identification

参考文献/References:


[1] 张颖, 潘静. 中国森林资源资产核算及负债表编制研究-基于森林资源清查数据[J]. 中国地质大学学报(社会科学版), 2016, 16(6): 46-53. ZHANG Y, PAN J. Research on asset accounting and statement formulation of forest resources in China based on forest resources inventory data[J]. Journal of China University of Geosciences(Social Science Edition), 2016, 16(6): 46-53. DOI:10.16493/j.cnki.42-1627/c.2016.06.005.
[2] 吴斌, 刘合翔. 人工智能技术在精准林业中的运用与发展[J]. 中国林业产业, 2006, 5: 25-28. WU B, LIU H X. Application and development of artificial intelligence technology in precision forestry[J]. China Forestry Industry, 2006, 5: 25-28.
[3] OTHMANI A A, JIANG C S, LOMENIE N, et al. A novel computer-aided tree species identification method based on burst wind segmentation of 3D bark textures[J]. Machine Vision and Applications, 2016, 27(5): 751-766. DOI:10.1007/s00138-015-0738-2.
[4] BERTRAND S, BEN A R, CERUTTI G, et al. Bark and leaf fusion systems to improve automatic tree species recognition[J]. Ecological Informatics, 2018, 46: 57-73. DOI:10.1016/j.ecoinf.2018.05.007.
[5] KUMAR N, BELHUMEUR P N, BISWAS A, et al. Leafsnap: a computer vision system for automatic plant species identification[C]//European Conference on Computer Vision. Berlin, Heidelberg:Springer, 2012: 502-516. DOI:10.1007/978-3-642-33709-3_36.
[6]瘙塁EKEROGLU B, INAN Y. Leaves recognition system using a neural network[J]. Procedia Computer Science,2016,102:578-582.DOI:10.1016/j.procs.2016.09.445.
[7] 穆易. 复杂背景下树干图像分割算法及其识别系统的研究[D]. 天津:天津理工大学, 2015. MU Y. Research on the segmentation algorithm and recognition system of trunk image in complex background[D]. Tianjin: Tianjin University of Technology, 2015.
[8] 于海鹏, 刘一星, 刘镇波. 基于图像纹理特征的木材树种识别[J]. 林业科学, 2007, 43(4): 77-81. YU H P, LIU Y X, LIU Z B. Wood species retrieval on base of image textural features[J]. Scientia Silvae Sinicae,2007,43(4):77-81.DOI:10.3321/j.issn:1001-7488.2007.04.013.
[9] 梁勇. 基于GLCM木材树种识别方法的研究[D]. 南京:南京林业大学, 2010. LIANG Y. The research of essence recognition method based on gray level co-occurrence matrix of wood digital image[D]. Nanjing: Nanjing Forestry University, 2010.
[10] 孟腊梅. 基于遗传规划的树皮纹理图像识别方法[D]. 保定:河北农业大学, 2011. MENG L M. Bark texture image recognition method based on genetic programming [D]. Baoding: Hebei Agricultural University, 2011.
[11] 陈勇平, 郭文静, 王正. 基于颜色直方图的木材单板图像检索技术研究[J]. 南京林业大学学报(自然科学版), 2015, 39(5):129-134. CHEN Y P, GUO W J, WANG Z. Image retrieval techniques of wood veneers based on color histogram[J]. Journal of Nanjing Forestry University(Natural Sciences Edition), 2015, 39(5): 129-134.DOI:10.3969/j.issn.1000-2006.2015.05.021.
[12] 窦刚, 陈广胜, 赵鹏. 采用颜色纹理及光谱特征的木材树种分类识别[J]. 天津大学学报(自然科学与工程技术版), 2015, 48(2): 147-154. DOU G, CHEN G S, ZHAO P. Wood species recognition using color texture and spectral features[J]. Journal of Tianjin University(Natural Science and Technology Edition),2015,48(2):147-154.DOI:10.11784/tdxbz201407025.
[13] 高程程, 惠晓威. 基于灰度共生矩阵的纹理特征提取[J]. 计算机系统应用, 2010, 19(6): 195-198. GAO C C, HUI X W. Texture feature extraction based on gray level co-occurrence matrix[J]. Computer System Applications, 2010, 19(6): 195-198. DOI:10.3969/j.issn.1003-3254.2010.06.047.
[14] 杨洋. 基于小波变换及SVM理论的树木种类识别研究[D].哈尔滨:东北林业大学,2017.YANG Y. Reseaech of tree species identification based on the theory of wavelet transform and SVM[D]. Harbin: Northeast Forestry University, 2017.
[15] 李可心, 戚大伟, 牟洪波, 等. 基于灰度共生矩阵与SOM神经网络的树皮纹理特征识别[J]. 森林工程, 2017, 33(3): 24-27. LI K X, QI D W, MU H B, et al. Identification of tree bark texture characteristic based ongray co-occurrence matrix and SOM neural network[J]. Forest Engineering,2017,33(3):24-27.DOI:10.3969/j.issn.1006-8023.2017.03.006.
[16] 赵鹏超, 戚大伟. 基于卷积神经网络和树叶纹理的树种识别研究[J]. 森林工程, 2018,34(1): 56-59.ZHAO P C, QI D W. Tree species recognition based on convolution neural network and leaf texture[J]. Forest Engineering, 2018,34(1): 56-59.DOI:10.16270/j.cnki.slgc.2018.01.012.
[17] HINTON G E. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786):504-507.DOI:10.1126/science.1127647.
[18] BENGIO Y, LAMBLIN P, POPOVICI D,et al. Greedy layer-wise training of deep networks[C]// Proceedings of the 2007 Advances in Neural Information Processing Systems. Vancouver: The MIT Press, 2007:153-160.DOI:10.7551/mitpress/7503.003.0147.
[19] RANZATO M A, POULTNEY C,CHOPRA S,et al.Efficient learning of sparse representations with an energy-based model[C]// Proceedings of the 2007 Advances in Neural Information Processing Systems. Vancouver: The MIT Press, 2007:1137-1144.DOI:10.7551/mitpress/7503.003.0147.
[20] ERHAN D, BENGIO Y, COURVILLE A, et al. Why does unsupervised pre-training help deep learning?[J]. Journal of Machine Learning Research, 2010, 11(3): 625-660. DOI: 10.1145/1756006.1756025.
[21] KLEMM D, KRAMER F, MORITZ S, et al. Nanocelluloses:a new family of nature-based material[J]. Angewandte Chemie International Edition, 2011,50(24):5438-5466.DOI:10.1002/anie.201001273
[22] YOO H J. Deep convolution neural networks in computer vision: a review[J]. IEIE Transactions on Smart Processing and Computing, 2015, 4(1): 35-43. DOI:10.5573/ieiespc.2015.4.1.035.
[23] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324. DOI:10.1109/5.726791.
[24] O'SHEA K, NASH R. An introduction to convolutional neural networks[EB/OL]. 2015: arXiv preprint arXiv:1511.08458, 2015.
[25] ZEILER M D,FERGUS R.Stochastic pooling for regularization of deep convolutional neural networks[C]//Proc of International Conference on Learning Representations.2013:1301-1310.
[26] PASSALIS N, TEFAS A. Learning bag-of-features pooling for deep convolutional neural networks[C]//Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 5755-5763. DOI:10.1109/ICCV.2017.614.
[27] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90. DOI:10.1145/3065386.
[28] ABADI M, BARHAM P, CHEN J M, et al. Tensorflow:a system for large-scale machine learning[EB/OL]. 2016:arXiv: 1605.08695[cs.DC].http://orxiv.org/abs/1605.08695.

备注/Memo

备注/Memo:
收稿日期:2018-09-07 修回日期:2019-03-28基金项目:国家重点研发计划(2017YFC0504106)。第一作者:刘嘉政(liujiazheng0919@163.com)。*通信作者:王雪峰(xuefeng@ifrit.ac.cn),研究员,ORCID(0000-0002-2703-9293)。
更新日期/Last Update: 2020-01-15