【作者】 杨雪清；

【导师】 黄心渊； 徐泽鸿；

【作者基本信息】 北京林业大学 ， 林业装备工程， 2008， 博士

【摘要】 森林资源作为人类赖以生存的物质基础,是国家自然资源的重要组成部分,在国民经济发展与建设中具有重要地位和作用。其中,森林资源调查已成为林业的重要基础工作。通过森林资源调查,可及时掌握森林资源数量和质量,从而为国家、地区及森林经营单位制定林业方针政策,编制林业区划、规划(森林经营方案),指导林业生产提供及时、准确的基础信息。森林资源规划设计调查(二类调查)是森林资源调查的主要内容,其主要成果图之一是林相图。林相图是指挥林业生产的地图,也是森林资源经营管理中使用最多的图形资料,但其调查频度却是十年一次,难以满足经济快速发展的需求,因此林相图的更新成为了极为重要的核心内容。传统的林相图更新方法大部分都是手动更新,时效性和准确性很难满足林业调查主管部门和森林资源管理工作的要求,遥感技术的发展为林相图更新提供了新的技术和手段。然而,利用遥感等信息技术实现林相图的计算机辅助更新在业务层面、技术层面和应用层面仍面临若干问题。从业务层面上看,虽然有不少研究成果,但是普遍是独立研究,难以整合,目前尚未形成一体化的林相图更新流程;从技术层面上看,影响林相图更新的核心分类方法尚有待进一步完善;从应用层面上看,目前尚缺乏将林相图更新过程中每一个环节进行集成,形成有界面、易操作、高效率的集成应用系统。鉴于此,本文在对发达国家的森林资源监测和管理水平研究的基础上,以林相图更新为研究对象,着重解决了我国森林资源遥感监测中的技术难点和关键问题,对于促进和提高我国森林资源监测与管理的信息化和现代化具有重要意义。在业务层面上,结合森林资源二类调查要求,提出了利用SPOT卫星数据对林相图更新流程;在技术层面上,对林相图更新的基础技术——森林资源图斑信息提取方面进行了深入研究,构建面向林相图更新的高效分类识别模型;在应用层面上,提出了森林资源数据库建设的标准和技术规范,对所有流程进行封装,研究开发功能较为完整、运行良好的林相图更新支持系统。本论文的主要内容包括:①研究如何根据森林资源二类调查规程,整合遥感图像预处理、遥感图像处理、森林资源图斑信息提取、小班数据更新、林相图生成等环节,形成有机集成业务流程、技术流程和质量控制流程的、基于SPOT5影像的一体化林相图更新流程;②研究如何利用林班线和山脊线及其他数据,对遥感影像数据的分类模型进行定制和增强,在缩短数据处理时间,提高分类效率的同时,保证分类精度;③研究如何采用面向对象方法和组件技术,封装遥感图像快速校正、融合以及森林资源分类、小班边界提取、专题图制作等实用功能技术,构建有界面、可视化、易操作的集成应用系统。本论文的主要创新点和贡献包括:①针对目前缺乏一体化的基于SPOT5的林相图更新流程问题,本论文首次提出了一个基于SPOT5遥感影像的林相图更新一体化流程,有机整合了业务流程、技术流程和质量控制流程,填补了目前林业领域的流程空白。该流程已在实际业务中进行推广应用,并取得了较好成果;②针对现有林相图分类方法中人工工作量大、分类处理时间长、分类结果准确率较低等问题,在传统分类方法的基础上,本论文提出了一个集成了无监和有监分类方法的混合分类模型,本混合模型首先通过利用林班线、山脊线等信息将SPOT5图像进行分块,避免分类算法在整个SPOT图像上的直接运算,降低了数据处理量,继而通过无监分类模型在像素级别将图形自动进行类别判断和聚合,再对聚合的图斑使用有监模型进行人工干预下的有监分类,避免了直接在像素级别的人工有监分类工作,减少了人工工作量的同时,也提高了分类的准确性。通过对比分析,新的混合模型提高了森林资源图斑信息提取的效率和准确性,解决了实现林相图更新从手工操作到半自动操作转变的重要技术问题;③针对目前我国缺乏可以实际运行的产业化的林相图更新应用系统问题,本课题在流程研究和技术攻关的基础上,将林相图更新中的数据处理、分类、分析等关键技术进行有机封装整合,设计开发了高效易用的林相图更新计算机辅助支持系统,为实现林相图从手动更新到半自动更新的转变提供了支撑,为提高林相图更新的工作效率提供了有力保障。更多还原

【Abstract】 Not only is forest a key component of national nature resource, but also is the physical foundation we live by, and it plays a critical role in national economic development. Through forest inventory, forest quantity and quality data can be obtained timely to support a nation, region, or organization with accurate base information in making forest policies, preparing forest division and management plans, and guiding forest productions. Forest inventory has become an essential work for forestry.Forest Management Inventory (Category II Inventory) is a major part of forest inventory. One of its key outcomes is the forest sub-compartment map, the base map guiding the forestry production and the graphic document most widely used in forestry operation management. Since Forest Management Inventory is performed every ten years, updating of the forest sub-compartment map become an important part of it. Most of the traditional updating methods are done manually, and their effectiveness and accuracy cannot satisfy the requirements of forest inventory administrative department as well as the forest management works. The development of satellite technology has provided the updating of forest sub-compartment map with new technology and instruments.As such, on the base of the researching of current development status of forest monitoring and management in developed countries, this paper focuses on the study of updating technology of forest sub-compartment maps, and mainly resolved the key technical issues and problems in forest monitoring with remote sensing. This is of great importance in promotion and improvement of the information and modernization in forest monitoring and management. On business process level, this paper proposed a process flow for the updating of forest sub-compartment map with SPOT 5 satellite data; on technical level, this paper studied the enabling technology in updating of forest sub-compartment map: the forest polygon information extraction technology, and constructed a efficient classification model for the updating of forest sub-compartment map; on the application level, this paper proposed standards and technical specifications for forest databases, packaged all the processes, studied and developed a well performed forest sub-compartment map updating support system which is fairly complete functionally.The main contents of this paper include:①Study how to follow the forest management inventory specifications and design a streamlined forest sub-compartment updating business process using SPOT5 imagery that seamlessly integrates the various technology processes in imagery pre-processing, processing, forest land information extraction, updating of sub-compartment data, and production of forest sub-compartment map;②Study how to utilize forest compartment border lines, mountain ridge lines, and other types of forest specific data in enhancing the SPOT imagery quality, to speed up the processing time and improve the classification efficiency while maintaining the accuracy;③Study how to integrate satellite imagery fast correction and fusing technologies, forest classification technology, sub-compartment border extraction technology, and forestry monographic map production technology to construct an efficient and effective forest sub-compartment updating support system using object oriented design philosophy and component based design principals;The major contributions of this paper include:④Addressing the lack of streamlined forest sub-compartment map updating business process issue, this paper proposes the first streamlined forest sub-compartment updating business process using SPOT 5 satellite imagery. This process seamlessly integrates business, technical and quality control processes and fills in the blanks in this field. It has been successfully applied it to the real-world business and is well acknowledged;⑤Addressing the issues of heavy manual involvements, long process time, and low accuracy issues in current forest classification modules, this paper proposes a hybrid classification model for updating of forest sub-compartment map that integrates supervised an unsupervised classification models. The model starts by partitioning the SPOT 5 imagery using forest borders, mountain ridges lines to avoid the calculation on whole SPOT 5 imagery and in turn to reduce the data to be processed. Then the model uses unsupervised classification to determine the categories of the pixels and aggregate them to form forest polygons. The aggregated polygons are then classified with expert inputs using supervised classification algorithms. This avoids the supervised classification on pixel level and thus reduces the involvements of experts without sacrificing the classification accuracy. Based on experimental comparison, this new hybrid model greatly improves the efficiency and accuracy of the classification, and clears one of the key issues in advance manual updating of forest sub-compartment map to computer aided updating;⑥Addressing the issue of lack of practical and industrialized forest sub-compartment map updating system, based on in-depth study of the business process and technical issues in this field, an effective and efficient computer aided forest sub-compartment map updating system has been designed and developed which seamlessly integrates various data processing, classification, and analysis models. The system can facilitate the transformation of forest sub-compartment map updating from manual to semi-automatic and in turn make it possible to greatly improve the updating efficiency.更多还原