【作者】 耿英；

【导师】 李淼；

【作者基本信息】 中国科学技术大学 ， 模式识别与智能系统， 2009， 硕士

【摘要】 农作物是人类生产和生活所必需的资源,在我国国民生产中也占有较大的比例。病虫害是农作物生产的重要制约因素,它能导致农作物大面积减产甚至绝收,影响农作物品质。因此,对作物病虫害种类的识别研究具有重要的现实意义和应用价值。传统的农作物病虫害诊断主要依靠人工目测方式,但这种方式存在一些问题:一方面农民凭自己经验判断,有可能出现误诊;另外一方面由于技术人员或者专家不能及时来现场诊断,造成病情的延误。而这些都可以借助于计算机图像处理和模式识别技术加以解决,因此我们希望建立图像识别系统来对作物病虫害进行识别。采用模式识别与图像处理的方法,用计算机软件来对农作物病害叶片进行分析,从而实现农作物病害的自动诊断。本文以黄瓜病害为例,主要工作总结如下:1.黄瓜病害图像预处理。主要包括图像裁剪、通道选择、图像平滑、阈值分割、轮廓提取和病斑提取等步骤。首先通过图像裁剪技术去除病害叶片的复杂背景,选择病斑显示最为清晰的蓝色通道,利用中值滤波方法对图像进行平滑,然后用阈值法分离病斑得到二值图像,最后用轮廓跟踪算法提取病斑轮廓并与原始图像进行叠加,得到病斑图像。2.病斑图像特征提取。对预处理后得到的病斑图像进行颜色、纹理和形状特征的提取,并将提取的特征按照SVM模型训练的格式要求保存在一个文本文件中。特征向量包括6种颜色特征、7种纹理特征和10种形状特征,共有23种特征。3.基于SVM方法的分类器设计与模型训练。将存放特征向量信息的文本文件交付分类器进行训练,得到黄瓜病害诊断模型。4.系统集成实现。使用Visual C++ 6.0和OpenCV开发了基于图像处理的黄瓜病害诊断系统CDRS 1.0,实现了黄瓜霜霉病、褐斑病、角斑病的快速识别。更多还原

【Abstract】 Crop production is necessary resource for human’s production and life and also a large proportion of our national product. Diseases and pests are important factors to restrict the growth of crops in agriculture producing, which may reduce yields of crops greatly and quality of products. Therefore, the research on identification of the type of crop pests and diseases has important practical significance and application value.At present, the diagnosis of crops diseases mostly depends on manual recognition, but some problems occur: on the one hand, it can be mistakenly diagnosed by farmers because they usually judge the symptom by their experiences; on the other hand, the disease treatment may be dallied over because the technician or expert can’t go to locale to diagnose in good time. All these can be resolved through computer image processing and pattern recognition technology. So we hope to build an image recognition system to identify diseases and pests of crops. Pattern Recognition and Image Processing by way of computer software used to analyze diseases on leaf of crop in order to achieve the automatic diagnosis. Cucumber disease leaf was as an example in this paper and the major work is summarized below:1. Pre-processing for disease image of cucumberPre-processing on image of cucumber diseases, which includes clipping, channel selecting, smoothing, segmentation, contours extraction and spots extraction. Firstly, we moved complex background for image with the image clipping technology and select blue channel on which the spots displayed most clearly, and then wiped noises for the image with Median filter. Secondly, threshold method was used to separate spots, and the outcome is a binary image. Lastly, the spots contours were extracted to plus with the original image, and then, the spots were extracted.2. Feature-extraction for spots imageColor, texture and shape features of the image after pretreatment were extracted and stored in a text file with the format which in accordance with SVM model training. In this paper, 26 features which include 6 color features, 7 texture features and 10 shape features were extracted.3. Designing classifier with SVMGetting the diagnosis model of cucumber diseases by designing classifier with SVM method and training the stored features.4. System-realizationVisual C++ 6.0 and OpenCV were used to develop the cucumber disease recognition system CDRS1.0 which realized quick identification of cucumber downy mildew, brown spot and angular leaf spot based on image processing.更多还原