基于深度学习的玉米叶片病害识别方法研究

Identification of Maize Leaf Diseases based on Deep Learning

玉米叶片病害是造成的玉米质量差、产量低主要原因之一。为了对玉米叶片病害进行快速准确识别,提出了基于ResNet(Residual Neural Network)深度学习网络对玉米病害识别的方法,采用ResNet作为玉米病害识别的主体模型,利用数据增强技术来扩充数据集,扩充后的数据集图片包括训6000张练集和1645张测试集,并使用预训练网络AlexNet、GooLeNet和ResNet进行识别玉米叶片病害的性能对比实验,研究发现在批量尺寸为32个和epoch次数为16时ResNet50获得最高的分类准确率为92.82%,优于传统机器学习算法。

In order to solve a series of problems such as poor corn quality and low yield caused by corn leaf disasters,a method based on the ResNet(Residual Neural Network)deep learning network to identify corn diseases is proposed.ResNet is used as the main model of corn disease identification,using data Enhanced technology to expand the data set.The expanded data set includes 6000 training sets and 1645 test sets.The pre-trained networks AlexNet,GooLeNet,and ResNet are used to perform performance comparison experiments for identifying corn leaf diseases.The research found that the batch size When the number of epochs is 32 and the number of epochs is 16,ResNet50 achieves the highest classification accuracy of 92.82%,which is better than traditional machine learning algorithms.