家畜饲料植物叶片的识别研究  

Research on Recognition of Livestock Feed Plant Leaves

作　　者：孟瑶 杨怀卿[1] 杨惠斐 杨华[1] Meng Yao;Yang Huaiqing;Yang Huifei;Yang Hua(College of Information Science and Engineering,Shanxi Agricultural University,Jinzhong Shanxi 030801)

机构地区：[1]山西农业大学信息科学与工程学院,山西晋中030801

出　　处：《中国农学通报》2021年第31期145-150,共6页Chinese Agricultural Science Bulletin

基　　金：国家自然基金“物联网温室环境控制系统随机模型建立及鲁棒控制研究”(31671571)。

摘　　要：为实现单一背景下可作家畜饲料的植物叶片实时自动识别,构建Alexnet卷积神经网络(Alexnet Convolution Neural Network)叶片识别模型。在山西农业大学植物园内,利用智能手机对8种可作家畜饲料的植物叶片进行拍摄,共计5130张,随机抽取4104张用于模型训练、1026张用于模型测试。采用灰度化处理、数据增强的方法对叶片图像进行预处理,构建Alexnet卷积神经网络模型,分别对预处理前和预处理后的植物叶片进行模型训练。测试结果表明,Alexnet卷积神经网络模型对预处理前的可作家畜饲料的植物叶片识别准确率为78.52%,对预处理后的可作家畜饲料的植物叶片识别准确率为98.38%。该模型能够较为准确地识别可作家畜饲料的植物叶片。To realize the real-time automatic recognition of plant leaves in a single background that can be used for livestock feed, in this paper, the Alexnet convolutional neural network model for leaf recognition was constructed. In the botanical garden of Shanxi Agricultural University, the mobile phone was used to collect 8 plant leaves that could be used for livestock feed, and a total of 5130 pieces were collected. Among them, 4104 pieces were randomly selected for model training, and 1026 pieces were used for model testing. The ways of gray-scale processing and data enhancement were used to preprocess the leaf image, constructed the Alexnet convolutional neural network model, and carried out model training on the plant leaves before and after the pretreatment, respectively. The test results show that the model has an accuracy rate of 78.52% for plant leaves before pretreatment and 98.38% for plant leaves after pretreatment. The proposed model can accurately identify the plant leaves.

关 键 词：家畜饲料 植物叶片 灰度化处理 图像增强 卷积神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]