【作者】 陈林；

【导师】 赵惠燕；

【作者基本信息】 西北农林科技大学 ， 农业昆虫与害虫防治， 2004， 硕士

【摘要】 确定靶标害虫的综合治理防治方案是一个包括害虫识别鉴定、种群与群落结构与动态分析、预测、合适的方法选择等复杂的综合决策过程。这个综合决策过程完成的质量和效率关系到农业生产者的增产增收程度，也体现着植物保护工作的现实意义。准确地识别害虫、开展预测预报、制定经济阈值、确定防治方案等等工作需要紧密的配合，才能实现良好的综合决策。由于这些工作之间相互关系复杂、要求大量的领域知识协同，普通农业生产者由于缺乏植物保护专业知识背景，在进行害虫鉴定的时候存在一定的困难；预测预报与防治方案的制订则更为困难。另一方面，各高等农业院校、研究单位在害虫识别、鉴定、决策等方面积累了大量的经验和知识，但是由于成果推广不利等因素的限制，不能够直接用于农业生产。随着人工智能技术，特别是专家系统的发展，为植物保护专业知识通过专家系统的方式方便地应用到农业生产领域提供了可能。专家系统拥有综合性的知识和高速处理知识的特点,且不受时间、空间的限制和人类情感的影响，能够方便、经济的应用到农业生产中。本文简要回顾了农业专家系统的发展、现状和趋势，分析了现有综合决策系统中出现的一些问题，如：鉴定子系统存在的鉴定方式要求强昆虫分类学背景基础、效率不高的问题，并提出了初步解决方案:强调生物学、生态学知识，降低使用时的知识背景要求；利用COM技术制作组件，实现分布式计算，降低更新难度；以框架结构组织知识库，降低维护成本；决策支持系统方面，本文实现了与鉴定专家子系统无缝连接，即将鉴定结果以及用户的提供的事实作为参数制作启动命令，启动决策支持系统进行推理，构造防治方案，提供给用户参考使用。本文主要利用Visual Studio 6套件中的Visual Basic，Visual InterDev及Macromedia MX中的Dreamweaver、Firework等工具进行了开发，以普通农业生产者、基层农技工作者为主要用户，实现了一支应用于农业生产的综合决策支持系统，在植物保护专业知识与计算机技术交叉领域做了一次有益的尝试，取得了如下的成果。本文分析对比了产生式系统、语义网络方式、框架式系统、面向对象方式、基于案例方式、基于模型方式六种知识表示方式，并结合植物保护专业领域知识，阐明了本文建立专家系统知识库使用的知识表示方法原因及应用原则。本文选择重要经济害虫设计并实现了鉴定专家系统框架式知识库和决策支持系统知识库。根据植物保护领域知识，结合专家系统数据库进行了推理机引擎设计。提出了以专业可用度、侧面属性值权重为主的冲突消解策略，并将其应用到推理机引擎中实现了推理引擎启发性的广度优先搜索策略，经本文试验，推理效果比较理想。系统集成方面利用COM技术开发设计了专家系统外壳以及专家系统推理机、专家系统插件工具等COM组件，构成了C/S/S三层结构的系统结构，实现分布式计算，提高了专家系统推理效率。以公布专家系统外壳接口、推荐专家系统推理机引擎接口等方法为二次开发的提供了基础，提高了本文设计的专家系统的可扩展性能。本文首次利用分布<WP=6>式技术构造实现了农业害虫鉴定专家系统、决策支持系统。另外，本文还设计了专家系统辅助工具若干：解释功能生成引擎能为配有解释演示性数据的文字描述提供动态网页概念解释；专家系统知识库制作工具为专家系统知识库的二次开发提供帮助并保证系统逻辑表的连接指向、逻辑结构正确。更多还原

【Abstract】 The basements of successful integrated pest management (IPM) are correct pest identification, accurate forecasting and rational controlling protocol and so on. Basing those knowledge and high-techs in the fields of computer and internet, Systems of Artificial Intelligence were developed rapidly.We simply reviewed the agricultural expert system, and analyzed some problems in expert system of insect identification, such as faint function, low efficacy and demand of more knowledge of taxonomy, etc. Aiming to those points, we brought solutions forward including depressing requirement of taxonomical knowledge while emphasizing employment on knowledge in biology and ecology. In order to achieve distributed calculation and facilitate updating of components in the expert system, we designed some others which are relatively independent using COM technique. We organized knowledge database in frame structure to decrease the cost of maintenance of database.Utilizing some powerful tools (such as Microsoft Visual Basic, Visual InterDev and Macromedia Dreamweaver, Firework), we established the expert system project which assist farmers and technician making decision in IPM. There are generally six organizing and expressing systems of knowledge in expert system which including Production System, Meaning Net System, Frames System, Object-Oriented System, Case Based Reasoning (CBR) and Model Based Reasoning (MBR). From the point of view of IPM, we compared those systems, chose two most adaptive systems to establishment of our expert system of controlling pests and elucidated the reasons and principles of combined use of Object-Oriented System and Frame System. In this paper, we designed Frame Knowledge Database of Pest Identification and Knowledge Database of Making-Decision and orient them to controlling of most important economic and agricultural pest species. At the same time, combining expert system database with IPM knowledge, we created an Expert System Reasoning Engine and selected a strategy of knot killed which is based on Professional Importance Quotient (PIQ) and Property Fuzzy Power (PFP). As we had expected, heuristic-wide-prior search strategies had been come true after PIQ and PFP were applied into the engine.Using COM technology, we exploited the COM components of expert system, including Expert System shell, Expert System reasoning machine and Expert System plug-in tools. Expert System’s three-layer C/S/S structure which was composed of those components can realize distributed computation and enhance the reasoning efficacy. Moreover, our Expert Systems (including Pest Identification System and Making-decision System) realized by <WP=8>distributed computation have two interfaces in the shell and engine, respectively, which improve the expandable property of the system.In addition, we also designed some accessorial tools of Expert System such as Exposition Building Engine (EBE) and Expert System Knowledge Database Modifying Tool (MT). EBE can provide dynamics HTML pages’ explanations to text information with explaining and demonstrating database. MT can assure correct data link in System Logic Table and stable logical structure, which help programmer make secondary development easily.更多还原