摘要
针对传统黄瓜病害识别方法中提取到的分类特征容易受病害叶片形态多样性、光照和背景影响的问题,提出了一种基于卷积神经网络的黄瓜病害识别方法,并建立了一个具有6种黄瓜病害的155 000多幅训练叶片图像数据库。根据病害叶片图像的复杂性,利用卷积神经网络从该数据库中自动学习黄瓜病害叶片图像的属性特征,再利用Softmax分类器进行分类。试验结果表明,与基于特征提取的传统病害识别方法相比,该方法的识别性能较高。
Focused on the problem of the traditional cucumber disease recognition methods that the extracted classifying features were more susceptible to diversity of the diseased leaf image,illumination and background,a cucumber disease recognition system was proposed based on convolutional neural network( CNN),and a cucumber disease leaf image database that contained more than 155 000 training images from six kinds of leaf diseases was established. Duo to the complexity of cucumber leaf images,the method of CNN was used to learn the recognition features adaptively from the database,and cucumber diseases were classified by Softmax classifier. Experimental results showed that the proposed method could achieve better performance in terms of classification than the traditional feature extraction based cucumber disease recognition methods.
出处
《江苏农业学报》
CSCD
北大核心
2018年第1期56-61,共6页
Jiangsu Journal of Agricultural Sciences
基金
国家自然科学基金项目(61473237)
河南省科技厅基础与前沿技术研究计划项目(172102210512
172102210510)
河南省教育厅高等学校重点科研项目(16A520095
16A510034)
关键词
黄瓜
病害识别
卷积神经网络
特征提取
Softmax分类器
cucumber
disease recognition
convolutional neural network(CNN)
feature extraction
Softmax classifier