摘要
传统植物图像识别研究主要集中在植物叶片图像。研究将深度神经网络学习运用于植物识别领域,突破局部叶片图像的限制,对常规植物图片进行识别。该方法运用google Net的深度卷积神经网络结构,通过图像旋转、镜像、随机裁剪等数据预处理方法扩充训练集,再利用SGD(随机梯度下降法)进行模型算法优化,生成对50种常规植物图像的识别模型。结果表明,该模型在测试集上能够达到平均90%的准确率。
Traditional plant image recognition research is mainly focused on plant leaf images.The deep neural network was applied to the field of plant recognition,it breaks through the restriction of the local leaf image and identifies the conventional plant pictures.GoogleNet deep convolution neural network structure was used, the training set was extend by data preprocessing methods such as image rotation, mirror image, random clipping and so on, and then SGD （stochastic gradient descent method ）was used to optimize the model algorithm to generate 50 kinds of common plant image recognition model.The results showed that the model could achieve an average accuracy of 90% on the test data set.
出处
《现代农业科技》
2017年第23期278-280,共3页
Modern Agricultural Science and Technology
关键词
植物图像识别
深度学习
神经网络
plant image recognition
depth learning
neural network
