基于深度可分离卷积神经网络的农作物病害识别方法
|
摘要
为了满足现代化、机械化农业生产的目标,降低模型的计算量,使农作物病害分类模型更适用于资源受限制的设备,提出了一种以深度可分离卷积为主的神经网络模型。利用深度可分离卷积和卷积相结合的方法取代标准卷积,计算量可降低至标准卷积的12%左右,并且大大减少网络模型的参数量。通过进一步减少通道数、改变网络输入图片大小的等方式,获得12种参数量和计算量不同的模型。结果显示,对含有复杂背景和光照不均匀的10类农作物的27种病害样本图片进行分类,该研究提出的模型准确率为98.26%,且参数量仅904 K。
In order to meet the goal of modern and mechanized agricultural production and reduce the computational load of the model,the crop disease classification model is more suitable for equipment with limited resources.A neural network model based on depthwise separable convolution was proposed.By combining depthwise separable convolution with 1×1 convolution,instead of standard convolution,the calculation amount could be reduced by 8～9 times,and the parameters of the network model could be greatly reduced.By further reducing the number of channels and changing the size of network input pictures,12 models with different parameters and computational load were obtained.Results showed that the accuracy of the proposed model was 98.26% and the parameters were only 904 K for the classification of 27 kinds of crop disease samples with complex background and uneven illumination.
In order to meet the goal of modern and mechanized agricultural production and reduce the computational load of the model,the crop disease classification model is more suitable for equipment with limited resources.A neural network model based on depthwise separable convolution was proposed.By combining depthwise separable convolution with 1×1 convolution,instead of standard convolution,the calculation amount could be reduced by 8～9 times,and the parameters of the network model could be greatly reduced.By further reducing the number of channels and changing the size of network input pictures,12 models with different parameters and computational load were obtained.Results showed that the accuracy of the proposed model was 98.26% and the parameters were only 904 K for the classification of 27 kinds of crop disease samples with complex background and uneven illumination.
引文
[1] 田凯,张连宽,熊美东,等.基于叶片病斑特征的茄子褐纹病识别方法[J].农业工程学报,2016,32(S1):184-189.
[2] 刘娜,赵慧,包思圆,等.基于图像处理的黄瓜叶子病害识别系统的研究[J].山东工业技术,2018(5):138.
[3] 田有文,李天来,李成华,等.基于支持向量机的葡萄病害图像识别方法[J].农业工程学报,2007,23(6):175-180.
[4] 杜海顺,蒋曼曼,王娟,等.一种用于农作物叶部病害图像识别的双权重协同表示分类方法[J].计算机科学,2017,44(10):302-306.
[5] SZEGEDY C,LIU W,JIA Y Q,et al.Going deeper with convolutions[C]//2015 IEEE conference on computer vision and pattern recognition (CVPR).Boston,MA,USA:IEEE Computer Society,2015.
[6] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Advances in neural information processing systems,2012,25(2):1097-1105.
[7] LIU W,ANGUELOV D,ERHAN D,et al.SSD:Single shot multiBox detector[C]//LEIBE B,MATAS J,SEBE N,et al.Compute vision-ECCV 2016.Zwizerland:Springer International Publishing,2016.
[8] REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:Unified,real-time object detection[C]//2015 IEEE conference on computer vision and pattern recognition (CVPR).Boston,MA,USA:IEEE Computer Society,2015.
[9] REN S,HE K,GIRSHICK R,et al.Faster R-CNN:Towards real-time object detection with region proposal networks[J].IEEE Trans Pattern Anal Mach Intell,2017,39(6):1137-1149.
[10] SCHROFF F,KALENICHENKO D,PHILBIN J.FaceNet:A unified embedding for face recognition and clustering[C]//2015 IEEE conference on computer vision and pattern recognition (CVPR).Boston,MA,USA:IEEE Computer Society,2015.
[11] 马浚诚,杜克明,郑飞翔,等.基于卷积神经网络的温室黄瓜病害识别系统[J].农业工程学报,2018,34(12):186-192.
[12] 龙满生,欧阳春娟,刘欢,等.基于卷积神经网络与迁移学习的油茶病害图像识别[J].农业工程学报,2018,34(18):194-201.
[13] MOHANTY S P,HUGHES D P,SALATHé M.Using deep learning for image-based plant disease detection[J].Frontiers in plant science,2016,7:1419.
[14] 孙俊,谭文军,毛罕平,等.基于改进卷积神经网络的多种植物叶片病害识别[J].农业工程学报,2017,33(19):209-215.
[15] HOWARD A G,ZHU M L,CHEN B,et al.MobileNets:Efficient convolutional neural networks for mobile vision applications[R].2017.
[16] CHOLLET F.Xception:Deep learning with depthwise separable convolutions[C]//2017 IEEE conference on computer vision and pattern recognition(CVPR).Honolulu,HI,USA:IEEE,2017.
[2] 刘娜,赵慧,包思圆,等.基于图像处理的黄瓜叶子病害识别系统的研究[J].山东工业技术,2018(5):138.
[3] 田有文,李天来,李成华,等.基于支持向量机的葡萄病害图像识别方法[J].农业工程学报,2007,23(6):175-180.
[4] 杜海顺,蒋曼曼,王娟,等.一种用于农作物叶部病害图像识别的双权重协同表示分类方法[J].计算机科学,2017,44(10):302-306.
[5] SZEGEDY C,LIU W,JIA Y Q,et al.Going deeper with convolutions[C]//2015 IEEE conference on computer vision and pattern recognition (CVPR).Boston,MA,USA:IEEE Computer Society,2015.
[6] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Advances in neural information processing systems,2012,25(2):1097-1105.
[7] LIU W,ANGUELOV D,ERHAN D,et al.SSD:Single shot multiBox detector[C]//LEIBE B,MATAS J,SEBE N,et al.Compute vision-ECCV 2016.Zwizerland:Springer International Publishing,2016.
[8] REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:Unified,real-time object detection[C]//2015 IEEE conference on computer vision and pattern recognition (CVPR).Boston,MA,USA:IEEE Computer Society,2015.
[9] REN S,HE K,GIRSHICK R,et al.Faster R-CNN:Towards real-time object detection with region proposal networks[J].IEEE Trans Pattern Anal Mach Intell,2017,39(6):1137-1149.
[10] SCHROFF F,KALENICHENKO D,PHILBIN J.FaceNet:A unified embedding for face recognition and clustering[C]//2015 IEEE conference on computer vision and pattern recognition (CVPR).Boston,MA,USA:IEEE Computer Society,2015.
[11] 马浚诚,杜克明,郑飞翔,等.基于卷积神经网络的温室黄瓜病害识别系统[J].农业工程学报,2018,34(12):186-192.
[12] 龙满生,欧阳春娟,刘欢,等.基于卷积神经网络与迁移学习的油茶病害图像识别[J].农业工程学报,2018,34(18):194-201.
[13] MOHANTY S P,HUGHES D P,SALATHé M.Using deep learning for image-based plant disease detection[J].Frontiers in plant science,2016,7:1419.
[14] 孙俊,谭文军,毛罕平,等.基于改进卷积神经网络的多种植物叶片病害识别[J].农业工程学报,2017,33(19):209-215.
[15] HOWARD A G,ZHU M L,CHEN B,et al.MobileNets:Efficient convolutional neural networks for mobile vision applications[R].2017.
[16] CHOLLET F.Xception:Deep learning with depthwise separable convolutions[C]//2017 IEEE conference on computer vision and pattern recognition(CVPR).Honolulu,HI,USA:IEEE,2017.