森林资源遥感调查综合处理与分析技术及业务化应用

针对国家森林资源连续清查（一类调查）与县级经营单位森林资源规划设计调查（二类调查）业务的重大需求，分别以中、高空间分辨率卫星遥感数据作为基础数据源，经过近十年的联合研究，首次建立了多阶遥感监测抽样技术体系，突破了森林资源遥感数据综合处理、分析及其集成应用的关键技术，规范了遥感技术应用的技术流程与标准，自主研发了森林资源调查遥感数据处理通用软件系统，建成了面向一类调查和二类调查两个服务层次的森林资源遥感监测业务应用系统，实现了森林资源遥感监测与信息管理的自动化、智能化和流程化。 项目的主要技术内容包括：①建立了基于GPS精确定位的全国284万个遥感样地与41.5万个固定样地相关联的森林资源遥感调查综合样地系统和遥感图像解译标志数据库，以及用于中分辨率遥感数据快速几何校正的控制点影像数据库 ②提出和建立了基于中等空间分辨率多光谱遥感影像解译标志数据库的森林类型自动分类、林分郁闭度与蓄积量分级估算模型，森林变化概率模型 ③建立了基于SPOT-5高空间分辨率遥感影像的小班边界提取、林相图及森林分布图快速提取更新技术 ④建立了基于激光雷达的林分平均高估算定量反演模型和方法 ⑤研发了具有自主知识产权、全组件化森林资源调查遥感数据处理通用软件系统和森林资源遥感信息定量提取与反演的专业化模块 ⑥通过系统集成，建立了基于自主遥感数据处理通用系统的国家级森林资源遥感监测和县级森林资源规划设计调查遥感业务应用系统。 项目成果的主要经济技术指标为：遥感影像批处理能力达到准实时，几何校正中误差优于0.5个像元 人机交互有林地判别正确率优于95%，针叶林、阔叶林、混交林和竹林等森林类型的判识正确率优于85% 业务化运行系统满足国家相关规范的要求，国家级森林资源遥感监测业务运行系统的运行成本比常规体系降低20% 县级森林资源规划设计调查的林相图和森林分布图的制作效率比常规提高2倍以上。 本项目共获得软件著作登记权8项 发表技术论文140余篇，其中SCI收录9篇，EI收录21篇，ISTP收录6篇 出版专著4部。其软件成果在生产中得到了很好应用。

To address the major demands of the National Forest Inventory (Type I Survey) and the County-level Forest Resource Planning and Design Inventory (Type II Survey) for forest resource management, this project took medium- and high-spatial-resolution satellite remote sensing data as the basic data source. After nearly a decade of joint research, it established the multi-level remote sensing monitoring sampling technology system for the first time, broke through the key technologies of comprehensive processing, analysis and integrated application of forest resource remote sensing data, standardized the technical procedures and standards for remote sensing technology application, independently developed a general-purpose software system for remote sensing data processing in forest resource surveys, and built forest resource remote sensing monitoring business application systems targeting the two service levels of Type I and Type II surveys, realizing the automation, intelligence and workflow standardization of forest resource remote sensing monitoring and information management.
The main technical contents of the project are as follows: ① Established a comprehensive forest resource remote sensing survey sample system and remote sensing image interpretation sign database that links 2.84 million national remote sensing sample plots and 415,000 permanent sample plots based on GPS precise positioning, as well as a control point image database for rapid geometric correction of medium-resolution remote sensing data; ② Proposed and established forest type automatic classification, stand canopy density and stock volume graded estimation models, and forest change probability models based on the interpretation sign database of medium-spatial-resolution multispectral remote sensing images; ③ Established technologies for rapid extraction and update of sub-compartment boundaries, stand type maps and forest distribution maps based on SPOT-5 high-spatial-resolution remote sensing images; ④ Established quantitative inversion models and methods for estimating stand average height based on LiDAR; ⑤ Developed a fully modularized general-purpose software system for remote sensing data processing in forest resource surveys with independent intellectual property rights, as well as professional modules for quantitative extraction and inversion of forest resource remote sensing information; ⑥ Through system integration, established a national-level forest resource remote sensing monitoring business application system and a county-level forest resource planning and design survey remote sensing business application system based on the self-developed general remote sensing data processing system.
The main economic and technical indicators of the project results are: the batch processing capability of remote sensing images reaches near-real-time; the root mean square error (RMSE) of geometric correction is better than 0.5 pixels; the accuracy rate of human-computer interaction-based forest land discrimination exceeds 95%; the recognition accuracy rates of forest types such as coniferous forest, broad-leaved forest, mixed forest and bamboo forest exceed 85%; the operational business system meets the requirements of relevant national specifications; the operating cost of the national-level forest resource remote sensing monitoring operational system is 20% lower than that of the conventional system; the production efficiency of stand type maps and forest distribution maps for county-level forest resource planning and design surveys is more than doubled compared with the conventional method.
This project has obtained 8 software copyright registrations, published more than 140 technical papers, including 9 papers indexed by SCI, 21 by EI and 6 by ISTP, and published 4 monographs. Its software achievements have been well applied in production.