基于改进卷积神经网络的多种植物叶片病害识别

发布时间：2018-05-24 10:29

本文选题：病害 + 植物　；参考：《农业工程学报》2017年19期

【摘要】：针对训练收敛时间长,模型参数庞大的问题,该文将传统的卷积神经网络模型进行改进,提出一种批归一化与全局池化相结合的卷积神经网络识别模型。通过对卷积层的输入数据进行批归一化处理,以便加速网络收敛。进一步缩减特征图数目,并采用全局池化的方法减少特征数。通过设置不同尺寸的初始层卷积核和全局池化层类型,以及设置不同初始化类型和激活函数,得到8种改进模型,用于训练识别14种不同植物共26类病害并选出最优模型。改进后最优模型收敛时间小于传统卷积神经网络模型,仅经过3次训练迭代,就能达到90%以上的识别准确率;参数内存需求仅为2.6 MB,平均测试识别准确率达到99.56%,查全率和查准率的加权平均分数为99.41%。改进模型受叶片的空间位置的变换影响较小,能识别多种植物叶片的不同病害。该模型具有较高的识别准确率及较强的鲁棒性,该研究可为植物叶片病害的识别提供参考。
[Abstract]:Aiming at the problems of long training convergence time and large model parameters, this paper improves the traditional convolution neural network model and proposes a new recognition model of convolution neural network combining batch normalization and global pool. In order to accelerate the convergence of the network, batch normalization of the input data of the convolution layer is carried out. The number of feature graphs is further reduced, and the global pool method is used to reduce the number of features. By setting initial layer convolution kernel and global pool layer type with different sizes, as well as setting different initialization types and activation functions, 8 improved models are obtained, which can be used to train and identify 26 kinds of diseases of 14 different plants and select the optimal model. The convergence time of the improved optimal model is less than that of the traditional convolution neural network model. After only three training iterations, the recognition accuracy of the improved model can reach more than 90%. The parameter memory requirement is only 2.6 MB, the average test recognition accuracy is 99.56, and the weighted average score of recall and precision is 99.41. The improved model is less affected by the change of the spatial position of leaves and can identify different diseases of many kinds of plant leaves. The model has high recognition accuracy and robustness, and it can be used as a reference for the identification of plant leaf diseases.
【作者单位】：江苏大学电气信息工程学院;江苏大学江苏省现代农业装备与技术重点实验室;
【基金】：国家自然科学基金资助项目(No.31471413) 江苏高校优势学科建设工程资助项目PAPD(苏政办发2011 6号) 江苏省六大人才高峰资助项目(ZBZZ-019) 江苏大学大学生科研立项资助项目(Y15A039);江苏大学大学生实践创新训练项目(No.46)
【分类号】：S432;TP391.41

【相似文献】