生态背景下基于人工智能深度学习的竹类害虫识别方法研究

编号 lyqk007617

中文标题 生态背景下基于人工智能深度学习的竹类害虫识别方法研究

作者 李禹辰  李非非  李见辉  余飞  徐杰 

作者单位 1. 电子科技大学 成都 611731;
2. 成都市森林病虫防治检疫站 成都 610032

期刊名称 世界竹藤通讯 

年份 2019 

卷号 17

期号 3

栏目名称 学术园地 

中文摘要 针对生态背景下的竹类害虫识别，作者研究了一种基于人工智能深度学习的识别方法。构建了具有5 663张图片的虫类数据集，其中包含3种竹类害虫和3种其他虫类，利用深度学习模型GoogLeNet特有的Inception模块构成的网中网结构，使其获得更多的图片特征，并开展了4组不同训练集与测试集比例的实验。结果表明：模型的精确度随训练集比重的增大而增大，当训练集和测试集的比例为9∶1时表现最好，F₁值达到了95.48%，模型精确度为97.5%，体现了识别模型具有较好的综合性能和较高的实用性。该方法能较好地实现3种竹类害虫在生态背景下的智能识别，是针对竹类生产经营中的虫害防治问题的一种智能化解决方案，为竹产业精细化管理及高效生产经营提供有效的科技支撑。

关键词 竹类害虫  虫类识别  人工智能  深度学习  生态背景 

基金项目 四川省科技支撑计划项目：“基于深度卷积特征的细粒度视觉图像识别模型与技术研究”（2018GZ0255）；“面向复杂环境的视觉目标检测与识别技术研究”（2019YFG0191）。

英文标题 Research on Bamboo Pests Identification Method Based on Artificial Intelligence and Deep Learning under the Ecological Context

作者英文名 Li Yuchen, Li Feifei, Li Jianhui, Yu Fei, Xu Jie

单位英文名 1. University of Electronic Science and Technology of China, Chengdu 611731, Sichuan China;
2. Chengdu Forest Pest Control and Quarantine Station, Chengdu 611731, Sichuan, China

英文摘要 This paper studied a bamboo pests identification model based on artificial intelligence deep learning under the ecological context. An insect dataset with 5 663 photos was established, consisting of 3 species of bamboo pests and 3 other insect species. Furthermore, a special network in network architecture composed of Inception modules in a deep learning model named GoogLeNet was introduced and applied to extract image features. Meanwhile, this model was run on 4 train-test set split experiments. The experimental results showed that the accuracy of this model increased with the increase of the ratio of the training set. When the ratio of the train-test set was set to 9:1, this model achieved the best performance, and when the mean-F1-Score reached 95.48%, the overall classification accuracy reached 97.5%. These experimental results implied that this model possessed relatively good comprehensive performance and practicability. This model could realize the intelligent recognition of the 3 species of bamboo pests mentioned above under the ecological context and is an intelligent solution to the pest control in bamboo production and management. Additionally, it could provide effective technology backup for intensive and effective management of bamboo industry.

英文关键词 artificial intelligence;deep learning;ecological context;bamboo pest;insect identification

起始页码 16

截止页码 21

作者简介 李禹辰(1997-),男,研究生,研究方向为人工智能与图像识别。E-mail:407676983@qq.com;李非非(1981-),男,高级工程师,硕士,长期从事林业调查规划设计与现代林业产业发展研究;E-mail:418223864@qq.com;余飞(1986-),男,工程师,长期从事林业调查规划设计工作;E-mail:15624285@qq.com

通讯作者介绍 徐杰(1981-),男,副教授,研究方向为人工智能与图像识别。E-mail:xuj@uestc.edu.cn

E-mail xuj@uestc.edu.cn

DOI 10.13640/j.cnki.wbr.2019.03.004

参考文献 [1] 周瑶. 基于机器视觉与黄板诱导的有翅昆虫统计识别系统的研究与实现[D]. 重庆:重庆大学, 2017.
[2] Klingenberg C P, Badyaev A V, Sowry S M,et al. Inferring developmental modularity from morphological integration:analysis of individual variation and asymmetry in bumblebee wings[J]. The American Naturalist, 2001, 157(1):11-23.
[3] 潘鹏亮, 史洪中, 尹新明. 昆虫数学形态学在桃红颈天牛雌雄成虫鉴别中的应用[J]. 河南农业科学, 2017, 46(12):159-164.
[4] 赵汗青, 沈佐锐, 于新文. 数学形态特征应用于昆虫自动鉴别的研究[J]. 中国农业大学学报, 2002, 7(3):38-42.
[5] Liu J D. The expert system for identification of tortricinae(Lepidoptera) using image analysis of venation[J]. Entomologia Sinica, 1996, 3(1):1-8.
[6] Liu J D. How to construct the expert system for species identification using venation of tortricinae(Lepidoptera)[J]. Entomologia Sinica, 1996, 3(2):133-137.
[7] Weeks P J D, O'Neill, Mark A, et al. Species-identification of wasps using principal component associative memories[J]. Image and Vision Computing, 1999, 17(12):861-866.
[8] Weeks P J D, O'Neill M A, Gaston K J,et al. Automating insect identification:exploring the limitations of a prototype system[J]. Journal of Applied Entomology, 1999, 123(1):1-8.
[9] Lim S, Kim S, Kim D. Performance effect analysis for insect classification using convolutional neural network[C]. 20177^th IEEE International Conference on Control System, Computing and Engineering (ICCSCE). Penang, Malaysia, 2017.
[10] 程尚坤. 基于深度学习的储粮害虫检测方法研究[D]. 郑州:河南工业大学, 2017.
[11] Cheng, X, Zhang, Y H, Chen, Y Q,et al. Pest identification via deep residual learning in complex background[J]. Computers and Electronics in Agriculture, 2017, 141:351-356.
[12] Motta D, Santos A Á B, Winkler I, et al. Application of convolutional neural networks for classification of adult mosquitoes in the field[J]. PloS One, 2019, 14(1):7-15.

PDF全文 浏览全文