面向粒度化环境的遥感可信计算研究

With the increasingly massive burden of remote sensing processing task, there is a need for high performance computing technology which may require both the much more stable computing environment and reliable processing algorithm. The application of trusted computing technique to concurrent computation can largely improve the controllability and reliability in computational process and results respectively. This study focus on particle-decomposing of the complex algorithms of remote sensing processing and, make sure these fractional pieces of algorithms run independently in the "data-computing" integrated concurrent computation platform. A trusted model is then introduced to ensure the reliability of the granular algorithms while the high computing efficiency is adequately maintained. For a complicated remote sensing computational task, which might be accomplished by conglomerating an array of granular algorithms in a way of "trusted chain", a self-adaptive mechanism is established to ensure the reliability of the assembled algorithms by the feedback information. That is, when a massive computing task is on-going, a temporal assessment result for the credibility of assembled algorithms is firstly calculated, and the trusted model is then auto-adjusted according to the assessment results until it is deemed to be acceptable, thus the reliability of computing results was ultimately guaranteed. As an initial application of trusted-computing technique to the field of remote sensing and geosciences, the specific theory and methodology of trusted-computing for remote sensing proposed in this study can meet the requirement of mass RS data processing in high performance computing platform in which efficient control for the processing procedure and computing results is of fundamental significance. In addition, it will also verify some of the significant concepts and approaches in trusted software or trusted network through actual application practice, and therefore will further promote the development of trusted computing theories.

遥感数据的处理规模日趋庞大，高性能计算技术的引入对计算环境的稳定性、处理算法的可靠性等都提出了更高要求，可信计算技术的应用可促进并行环境中计算过程的可控制及计算结果的可信赖。本项目重点研究将复杂遥感算法进行粒度化分解并独立运行于"数据-计算"一体化的并行计算环境，在充分发挥计算高效性的同时通过建立可信模型来保证算法的可靠性，并以可信链形式聚合多个粒度化算法来完成复杂的遥感计算功能，通过可信度评价保证结果的可信性，进而自适应地根据中间结果的可信度调整信任传递流程，最终提高计算结果的可信度。作为可信计算技术在遥感地学领域中的首次应用，遥感可信计算理论与方法的提出，既满足海量遥感数据在高性能计算环境下对处理过程和计算结果的有效控制需求，同时也对可信软件、可信网络中诸多重要概念与关键方法进行了真实实践验证，将促进可信计算理论的发展。

项目按照计划顺利进行，已经研究了面向海量遥感数据的高可信计算方法；收集整理了多源遥感影像以及部分用于测试的无人机影像，初步面向大数据计算解决了数据存储、传输等相关技术，以此作为关键技术成果，以测试方式，参与国家863计划项目“高性能GIS 关键技术与软件系统”；在项目资助下，发表论文4篇，其中SCI 2篇，国内核心期刊1篇。