复杂背景茶树病害图像识别  

Image Recognition of Tea Tree Disease in Complex Background

作　　者：余文森 YU Wensen(School of Mathematics and Computer Science,Wuyi University,Wuyishan,Fujian 354300,China)

机构地区：[1]武夷学院数学与计算机学院,福建武夷山354300

出　　处：《武夷学院学报》2022年第3期33-40,共8页Journal of Wuyi University

基　　金：福建省自然科学基金面上项目(2019J01833)。

摘　　要：针对复杂背景茶树病害图像样本收集困难问题,提出利用相似植物同种病害的图像增广数据集的新思路,构建小样本茶树病害图像数据集和增广数据集,同时引入网上公开的Plant Village植物病害图像数据集。在此基础上,采用VGG-16深度神经网络模型,通过实验比较分析迁移学习、数据增广、模型微调等小样本学习方法对复杂背景茶树病害图像识别问题的有效性。研究结论可为后续复杂背景茶树病害图像识别问题研究提供一定的借鉴和参考。Aiming at the difficulty of collecting tea tree disease image samples in complex background,a new idea of using images of the same disease of similar plants to augment the data set was proposed.The small sample tea tree disease image data set and the augmented data set were constructed,and the PlantVillage plant disease image data set,which was publicly accessible online,was introduced.On this basis,VGG-16 deep neural network model was used to compare and analyze the effectiveness of transfer learning,data augmentation,model fine-tuning and other small sample learning methods for image recognition of tea tree disease in complex background.The conclusion can provide reference for the subsequent research on image recognition of tea tree disease in complex background.

关 键 词：复杂背景 茶树病害图像识别 小样本学习 迁移学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]