【作者】 张艳荣；

【导师】 王霓虹；

【作者基本信息】 东北林业大学 ， 森林工程， 2012， 博士

【摘要】 森林作为一种人类社会发展过程中的最宝贵的战略资源,正在不断地遭受着各种灾害的侵蚀。在这些重大的森林资源灾害之中,森林病虫害这种频发性的生物灾害,一直都是林业发展的一个重要的制约因素。然而,我国森林病虫害总的发生又随着人工造林面积的持续增长而不断地上升。因此,应采取积极的应急措施,来减轻或避免森林病虫害所造成的损失。随着对“泛在林业”的要求不断地提高,使现代林业工作更加信息化、网络化和智能化,本文将粗糙集理论、专家系统、网络技术、现代信息技术以及领域专家的病虫害诊断、预测预报及防治经验和技术成果结合起来,应用在森林病虫害的预测、诊断和防治之中,从而对减少病虫害对森林的危害,对推动森林病虫害管理的现代化、科学化,实现信息化跨越式发展,对推进示范地区的森林常见病虫害监测、预警、预防、防治的体制和机制创新起着重要作用。为了能够准确地对森林病虫害进行预测预报,本文分析了影响森林病虫害发生发展的一些关键性因素,如林分结构、气候变化、生物因素、土壤因素和人为活动等,详细地分析了温度、湿度、降水量、光和风等气息因子及其对森林病虫害的发生发展的影响,并引用了温度和太阳辐射量的模拟模型以及湿度和降水量的模拟模型,利用这些气象模拟模型可得出该季节内逐日的最高温度、最低温度、降水量等气象因子的模拟数据,这有利于提高森林病虫害预测预报的准确度。为了能够及时地掌握森林病虫害种群数量变化的规律,对森林病虫害未来的发生状况以及增长趋势作出准确、科学的预测预报,以能够及时提出森林病虫害的防治措施,减少森林病虫害的危害造成的损失,本文应用发育进度预测法、有效积温预测法等对示范地区常见的森林害虫的发生期进行预报预报,并建立了相应的预测预报模型；应用有效基数预测法、数理统计预测法等对示范地区常见的森林害虫的发生量进行预报预测预报,并建立了相应的预测预报模型：利用马尔柯夫过程分析,对示范地区常见的森林病害的发病程度进行预测预报；利用灰色模型对示范地区常见的森林病害的发病过程的模拟和发病时点的预测；利用非线形模型对示范地区常见的森林病害的发展阶段进行预测预报；利用灰度模型分析,对示范地区常见的森林病害的感病指数进行预测预报。由于森林病虫害预测预报专家系统包含的知识量比较大,为了能够快速、准确地从杂乱无章的海量数据中挖掘潜在的有利用价值的信息并用于森林病虫害的预测预报,本文将粗糙集理论应用到了森林病虫害的预测预报过程之中。通过对森林病虫害预测预报的数据进行收集、完备化和离散化,提出了一种改进的基于差别矩阵的属性约简算法,基于此算法对森林病虫害预测预报条件属性集进行约简,从而对产生的规则进行提取与约简,得出了一种新的基于粗糙集理论的森林病虫害预测预报模型。本文采用三层B/S结构和基于J2EE标准的开发模式建立森林病虫害预测预报专家系统。这种模式可以使前端的表现与应用逻辑、数据存储相分离,通过组件的开发与部署策略,使整个系统的架构清晰灵活,方便部署和扩展。业务处理用Jsp+Javabean的方式实现并通过JDBC的方式访问数据层的数据资源。森林病虫害预测预报专家系统主要实现对示范地区常见森林病虫害的发生量与发生期的预测预报及损失评估,除此之外,该专家系统还包括树种信息、病虫害诊断、病虫害防治及病虫害查询等辅助模块。综上所述,本文的主要贡献如下：(1)粗糙集理论与人工智能技术首次应用于林业病虫害领域,取得了良好效果；(2)提出了示范地区常见的森林病虫害的发生期预测预报与发生量预测预报模型,通过对模型进行验证与精度分析,得出的模型准确度较高,且预测预报的结果与实际情况基本相符；(3)提出了一种改进的基于差别矩阵的属性约简算法,基于此算法对森林病虫害预测预报条件属性集进行约简,从而对产生的规则进行提取与约简,建立了一种新的基于粗糙集理论的森林病虫害预测预报系统,在试验示范区得到良好应用。本论文的研究成果可为基层森工企业以及广大的林农用户在森林病虫害的预报与防治的过程中提供理论指导和技术支持,为“泛在林业”建设提供了完整范例,同时也对农业病虫害预测预报与防治等专家系统的建立具有一定的指导意义和借鉴价值。更多还原

【Abstract】 As the most precious strategic resources of human society development, is constantly suffer from huge loss of various disasters. However in these major forest resources, the frequent biological disaster such as forest diseases and pests has been to limit the forestry development. With the artificial afforestation area continues to grow, forest diseases and pests is always in upward trend. Therefore, we should take active measures to reduce the loss and avoid the damage caused by forest pest and disease.With the wisdom of forestry increasing demands continuously, in order to make the modern forestry informationization, networking and intelligent, the rough set theory, expert system, network technology, modern information technology and experts in the field of forest diseases and pests diagnosis, prediction and control experience and technology combine, application in forest diseases and pests forecast, diagnosis and prevention, so as to reduce harm caurse by forest diseases and pests. It promotes the modernization, scientific of the management of forest diseases and pests. And it makes the realization of the development with the type of span informatization. This plays an important role to advancing demonstration areas of forest diseases and pests monitoring, early warning, prevention, control system and mechanism innovation.In order to forecasting of forest diseases and pests occurrence accurately, this paper analyzed the key factors of effect the forest disease and pests occurrence, such as forest structure, climate change, biological factors, soil factors and human activities, and make a detailed analysis of the factors of the temperature, humidity, rainfall, light, wind, and the influence of thess effects on forest disease and pestd occurrence and development. This paper cited the temperature and solar radiation simulation model and the humidity and precipitation simulation model, the simulation model of the seasonal weather can be obtained within the daily maximum temperature, minimum temperature, precipitation, meteorological factor of the simulated data, which is conducive to make the forecast accuracyIn order to timely control of forest diseases and pest population vari(?)tion of forest diseases and pests, the future situation and growth tendency to make scientific, accurate forecast, so as to give forest diseases and pests control measures, reduction of forest pests and diseases damage and loss. This paper applises developmental schedule forecasting method, effective accumulated temperature prediction method, regression analysis on demonstrative area common forest pest occurrence period forecast, and the establishment of the corresponding prediction model, application of effective base prediction method, mathematical and statistical method of demonstration area common forest pest occurrence quantity forecast, and the establishment of the corresponding prediction model, using Markov process analysis, demonstration area of common forest disease severity prediction using grey model for demonstration area, common forest disease process simulation and onset time prediction by using the non linear model, demonstration area of common forest disease stage forecasting, gray model analysis on utilization, demonstration common area of forest disease index predictioy.Forest diseases and pests forecasting expert system includes large quantity knowledge. In order to mining potentially valuable information from the mass data quickly, accurately and uses for the prediction of forest pests and diseases. This paper uses the rough set theory and applises it to the process of the prediction of forest pests and diseases. Through the forecast data are collected, complete and discretization, put forward a kind of improved attribute reduction algorithm based on discernibility matrix. Based on this algorithm on forest diseases and pests forecasting condition attribute set is reduced, thereby to generate rules extraction, obtains forest diseases and pests forecasting model based on a new rough set theory.This paper adopts three layers of B/S structure and based on the J2EE standard development mode for the establishment of forest diseases and pests forecasting expert system. This model can make the separation of the performance of the front and the application logic, data storage phase. It makes the whole system structure clear and flexible, convenient deployment and expansion through the component development and deployment strategy. Business processing with Jsp+Javabean mode and accessing data resources from data layer through the way of JDBC. Forest diseases and pests forecasting expert system based on demonstration area common forest diseases and pests occurrence period and amount of prediction and assessment of loss. In addition to this, the expert system also includes species information, diagnosis of diseases and pests, plant diseases and pests, diseases and pests query and other auxiliary module.In summary, this paper’s main contribution is as follows:(1) Rough set theory and artificial intelligence technology was first used in the field of forestry diseases and pests;(2) Proposed the model of forestry diseases and pests forecasting in the demonstration. Through the model validation and accuracy analysis, the accuracy of the result is high, and forecast results with the actual situation are match;(3) Proposed an improvement attribute reduction algorithm based on discernibility matrix. Reduced conditional attribute set of forestry diseases and pests forecasting based on the algorithm. Through the extraction and reduction of generation rules, obtained a new f forestry diseases and pests forecasting model based on rough set theory. By verifying, the model has achieved a good result. The research results of this paper can provides the theoretical guidance and technical support for the vast number of users and farmers to prevent the forest diseases and pests. And provides a complete sample for the digital forestry construction. It has the certain instruction significance and reference value on the expert system of forecast and prediction of agricultural pests and diseases.更多还原