【作者】 赖军臣；

【导师】 李少昆；

【作者基本信息】 石河子大学 ， 作物栽培学与耕作学， 2010， 博士

【摘要】 本文以病症图像信息为主线,以满足用户多样化的需求为导向,探索病症图像信息求解疾病的不同创新表达方法。综合运用数字图像处理技术、模糊逻辑理论、模式识别等领域技术,研究实现了玉米叶斑病害的自动诊断系统；运用专家系统、人工智能、网络信息技术,研究实现了基于图像规则的玉米病害诊断专家系统。主要内容包括：1.大田非结构环境下玉米叶斑病害分割算法的筛选。先用中值滤波算法去除图像噪声,然后利用拉普拉斯图像锐化算法增强图像。在图像增强的基础上,本文研究实现了基于领域的对噪声不敏高斯模型的Markov随机场分割模型(G-MRF)和区域增长分割算法(SRG),分割玉米叶斑病害,并与有代表性的作物病害分割算法相比较,优选了区域增长分割算法为玉米病害的自动化识别服务。2.玉米叶斑病害图像特征提取和优选。对预处理后的病害图像提取病斑的颜色、纹理、几何形状三方面14个特征值,再对这些特征分析计算后,最终优化和选择了似圆度、偏心率和矩形度3个最具有代表性的特征参数。3.玉米叶斑病害图像的自动分类识别。利用模糊模式识别对玉米叶斑病害图像进行分类,研究对比了隶属原则识别法和择近原则识别法,设计了相应的分类器,并建立了病害诊断模型。利用VC6.0作为开发工具,编写了“玉米病斑智能识别系统”,实现了对玉米常见叶斑类病害图像的处理与识别。4.基于图像规则的玉米病害诊断专家系统的构建。在对玉米病虫草害诊断与防治知识进行深入分析的基础上,将植保专业知识与用户的田间实际体验相结合,通过优化的人机交互界面,将推理规则以典型图像加通俗文字描述的直观形式呈现给用户。面向基层农技人员和农民用户,以及基于网络和单机软件等多种服务形式,解决农户的实际应用问题,研制开发了以田间典型特征事实图像作为人机互动形式,并具有界面友好、扩展性佳、实用性强的“玉米病害诊断专家系统”,并通过国家玉米产业体系在全国范围内推广应用。更多还原

【Abstract】 The paper took maize disease image for main materials, variety of users’requirements for guide to explore new solutions for identifying diseases with disease images. This study investigated technologies of corn leaf spot diseases automatic diagnosis by using digital image processing, fussy theory and pattern recognition, and constructed the’Expert System For the Diagnosis of Corn Diseases’ by using expert system, artificial intelligence and Network Information technology, The main contents are as followings:1. Selection of Image segmentation Algorithms. Algorithm of Middle filter was used to eliminate noise, and algorithm of image sharpening was used to enhance images. Afterward, Gauss-MRF model and seeded regional growing algorithm were realized and introduced to segment corn spot disease images, and these algorithms were compared with representative algorithm used to segment crop disease image in the past. Finally seeded regional growing algorithm was selected.2. Character extraction and selection. Primarily extracted 14 characteristic value from monochrome color, texture and geometry shape on Processed disease image, and then optimized and selected these characteristic value, finally got circularity, eccentricity and rectangle 3 representative characteristic parameter.3. Diseases automatic identification and diagnosis. Use fuzzy pattern recognition to classify the corn disease images, studied membership principle identification and closest principle identification, designed corresponding classifier, and built disease diagnostic model. Took Visual C 6.0 as the tool, compiled the program code, finally feasible software was formed.4. Construction of corn diseases diagnosis expert system based on image rule. Based on in-depth analysis of diagnosis and preventive treatment knowledge of corn diseases, the authors have developed the ’Expert System for Diagnosis of Corn Diseases’with friendly interface, easy expandability and practicability, which is designed as specific to agro-technicians at grass-root level and farmers users on the ground that it features human-computer interaction on actual images of typical characteristic facts in the fields, provides multi-service on the basis of both network system and stand-alone software. The system has been expanded and applied in the national scope through National Corn Industry Technique System.更多还原