基于卷积神经网络的葡萄叶片氮含量识别方法

Identification Method of Nitrogen Content in Grape Leaves Based on Convolutional Neural Network

中文关键词:  卷积神经网络  葡萄  图像识别  深度学习

英文关键词:convolutional neural network  grape  image recognition  deep learning

中文摘要:

      为实现对葡萄叶片氮素含量快速、便捷的识别，在卷积神经网络VGG-16网络结构基础上，将数据增广后的图像按不同梯度划分进行模型训练，通过十折交叉验证法探究最佳的训练集与验证集分配比例，并构建4个不同深度的网络模型进行训练对比，采用全局平均池化代替全连接层约简网络参数量。训练结果表明，氮含量梯度设为0.70%、0.35%和0.175%时，室内简单背景识别准确率分别为85.9%、76.2%和71.1%；晴天室外复杂背景下识别准确率分别为44.6%、35.0%和30.4%。研究结果表明利用VGG-16建立的网络学习模型对葡萄叶片氮含量识别提供了一种新的便捷方法，对农业信息化和智能化技术应用具有一定促进作用。

英文摘要:

      In order to rapidly and conveniently identify the nitrogen content in grape leaves，based on the VGG-16 network structure,the images are divided into four different gradients for model training.Then the training set and the verification set are divided into four ratio combinations to construct four different depths network models.Simultaneously,the global average pooling is used to instead the full connectivity layer.The training results show that when the nitrogen content gradients is 0.70%,0.35% and 0.175%,the average recognition accuracy under simple indoor background is 85.9%,76.2% and 71.1%,respectively,and the outdoor background in 44.6%,35.0% and 30.4%.The research results show that the model can identify the nitrogen content in grape leaves,which has certain promotion to the application of agricultural information and intelligent technology.