基于深度稀疏自编码网络的植物叶片分类

Plant leaf classification based on Softmax regression and K deep sparse autoencoder network

针对传统机器学习方法对植物叶片图像分类识别率不高的问题,探讨了基于深度稀疏自编码网络(Deep Sparse Autoencoder Network,DSAN)的植物叶片分类研究。自动编码器通过编码和解码重构输入数据,对植物叶片进行分层特征学习,在自动编码器上添加稀疏限制,对隐含层神经元进行压缩,从而学习到更高层的隐含特征用于分类,解决了因选取的特征表达不足导致网络模型分类性能不佳的问题。实验采用公开的植物叶片图像数据库MalayaKew(MK)作为研究对象,该数据集包含44类植物。将预处理之后的叶片图像直接作为输入数据,通过DSAN学习到叶片的高层特征,结合Softmax分类器用于分类。实验结果表明,该算法能够有效提高植物叶片图像的分类精度,在植物分类领域具有一定的应用价值。

Aiming at the low recognition rate of traditional machine learning methods for plant leaf image classification,the research of plant leaf classification based on Deep Sparse Autoencoder Network(DSAN)was discussed.Auto-encoder reconstructs input data by encoding and decoding,learns hierarchical features of plant leaves,adds sparse restrictions to the auto-encoder and compresses hidden layer neurons,thus learns higher hidden features for classification,and solves the problem of poor classification performance of network model caused by insufficient expression of selected features.In this paper,Malaya Kew(MK),a public image database of plant leaves,is used as the research object.The data set contains 44 kinds of plants.The pre-processed blade image is directly used as input data,and the high-level features of the blade are learned by DSAN,which is combined with the Softmax classifier for classification.The experimental results show that the algorithm can effectively improve the accuracy of plant leaf image classification,and has a certain application value in the field of plant classification.