| 基于多层感知神经网络的水稻叶瘟病识别方法 |
| |
| 引用本文： | 刘立波,周国民.基于多层感知神经网络的水稻叶瘟病识别方法[J].农业工程学报,2009,25(13):213-217. |
| |
| 作者姓名： | 刘立波 周国民 |
| |
| 作者单位： | 1．中国农业科学院农业信息研究所，北京 100081； 2．宁夏大学数学计算机学院，银川 750021;中国农业科学院农业信息研究所，北京 100081 |
| |
| 基金项目： | 国家高技术研究发展计划（863）资助项目（2007AA102Z237） |
| |
| 摘 要： | 为实现水稻叶瘟病的快速诊断，综合利用图像处理技术和神经网络来进行叶瘟病斑的识别研究。该文设计了3个多层感知分类器来进行病斑识别准确率的对比验证，分别采用叶片正常区域和病斑区域的纹理特征、颜色特征以及纹理和颜色的组合特征作为不同分类器的输入单元；输出层采用1个单元用于输出病斑区域和正常区域的识别结果。首先，该文将采集到的RGB图像转换成灰度图像，利用灰度共生矩阵分别提取叶片正常区域与病斑区域的能量、对比度、熵、逆差距作为纹理特征；紧接着，将RGB彩色空间转换至HIS和Lab空间，分别提取病斑区域和正常区域的L、a、b值作为颜色特征。最后，采用不同的BP神经网络分类器进行病斑区域识别。该文共采用120副图像作为待测对象，试验结果表明，采用颜色和纹理的组合特征进行识别，准确率要比单独使用纹理特征和颜色特征高10%～15%。本文的研究结果为进一步实现水稻病害自动诊断打下了基础。
|
| 关 键 词： | 多层感知器，特征提取，分类器，图像识别，水稻叶瘟病 |
| 收稿时间： | 2009/6/30 0:00:00 |
| 修稿时间： | 2009/9/11 0:00:00 |
Identification method of rice leaf blast using multilayer perception neural network |
| |
| Liu Libo and Zhou Guomin.Identification method of rice leaf blast using multilayer perception neural network[J].Transactions of the Chinese Society of Agricultural Engineering,2009,25(13):213-217. |
| |
| Authors: | Liu Libo and Zhou Guomin |
| |
| Institution: | 1. Institute of Agricultural Information, Chinese Academy of Agriculture Sciences, Beijing 100081, China; 2. College of Mathematics and computer, Ningxia University, Yinchuan 750021, China and 1. Institute of Agricultural Information, Chinese Academy of Agriculture Sciences, Beijing 100081, China |
| |
| Abstract: | In order to achieve the rapid diagnosis of rice leaf blast, authors utilized image processing techniques and neural network comprehensively for the recognition of leaf blast lesion. For the accuracy rate of comparative analysis on lesion identification, three multilayer perception classifiers were designed. Three features of normal and lesion part which were texture, color and combined feature of texture and color were selected as input unit for different classifier respectively. Output layer adopted one unit for the identify results of lesion and normal region. First of all, grayscale image was transformed from RGB image, using gray level co-occurrence matrix to extract energy, contrast, entropy, inverse gap as texture features from leaf lesion region and normal region respectively. The RGB color space was transformed to HIS and Lab space, and L, a, b values were extracted as color features from lesion area and normal region respectively, using different BP neural network classifier to identify lesion region.120 images were selected as test objects. The experimental results showed that if the combined feature of color and texture was used as input parameters, the accuracy rate would be 10%-15% higher than that of texture features and color features alone. The results of this paper laid a foundation for realization of the automatic diagnosis of rice diseases. |
| |
| Keywords: | multilayer neural networks feature extraction classifiers image recognition rice leaf blast |
|
| 点击此处可从《农业工程学报》浏览原始摘要信息 |
| 点击此处可从《农业工程学报》下载免费的PDF全文 |
|
