基于物联网和深度卷积神经网络的冬枣病害识别方法

发布时间：2018-04-28 00:00

本文选题：冬枣病害识别 + 冬枣病害图像　；参考：《浙江农业学报》2017年11期

【摘要】：针对传统的作物病害识别方法中人为提取的分类特征,对复杂作物病害图像的形状、光照和背景比较敏感等问题,提出一种基于物联网和深度卷积神经网络(DCNN)的冬枣病害识别方法。DCNN由1个输入层、4个卷积层、3个下采样层、1个全连接层和1个输出层组成。利用该方法能够提取冬枣病害图像的有效特征,并识别病害类型,避免了传统作物病害识别方法中繁琐的特征提取过程。在4种冬枣病害果实数据库上进行了冬枣病害识别实验,识别率达到92%以上。试验结果表明,该方法适合利用物联网采集的大规模视频病害图像进行冬枣病害识别。
[Abstract]:In view of the classification characteristics extracted by traditional crop disease identification methods, the shape, light and background of complex crop disease images are sensitive, and a new identification method based on Internet of things and deep convolution neural network (DCNN).DCNN is proposed, which consists of 1 input layers, 4 coiling layers, 3 down sampling layers, and 1 full connections. This method can extract the effective features of the winter jujube disease image and identify the disease type, and avoid the complicated feature extraction process in the traditional crop disease identification method. The experiment of identification of winter jujube disease on the 4 winter jujube disease fruit database is carried out, the recognition rate is up to 92%. The test results show that this method is more than 92%. The method is suitable for the identification of winter jujube disease by using large-scale video disease images collected by Internet of things.

【作者单位】：西京学院信息工程学院;
【基金】：国家自然科学基金项目(61473237) 陕西省自然科学基础研究计划(2016GY-141)
【分类号】：S436.65;TP183

【相似文献】