【作者】 石绥祥；

【导师】 于戈；

【作者基本信息】 东北大学 ， 计算机软件与理论， 2005， 博士

【摘要】 实现数字海洋需要空间上密集分布、时间跨度连续的各类海洋数据信息作为支撑。然而，由于海洋本身的广阔性，决定了信息获取的难度，致使我们无法获得海洋上任意点的真实信息；又由于目前我们所拥有的海洋观测手段的局限性，致使数字海洋中不同渠道获取的数据之间存在着许多的不确知性甚至冲突，这缘于许多因素，例如，测量误差、间接信息计算偏差等。因此，建设数字海洋系统将无法回避海洋多源数据的融合问题，特别是多渠道不确知性信息的融合。本文的重点就是要找到一种新的融合策略以解决这个问题。 尽管海洋领域学者已经开始探讨信息融合技术在海洋中的应用，许多文献也给出过部分成功的应用实例，但总体上仍处于起步阶段，仅限于具体算法研究与应用，系统性研究甚少。溯其根源，主要是因为海洋学领域的专业性成为信息融合学者的门坎儿，而海洋学工作者在信息科学和信息融合技术上的薄弱也限制了该技术在海洋领域的推广。 为解决数字海洋中多渠道不确知性信息的融合问题，作者进行了比较系统的研究，并取得如下进展： 首次给出了针对数字海洋的信息融合层次与方式，提出了数据层——应用层——服务层——顾问层四层融合模型，对数字海洋系统融合信息的粒度、特点和方式进行了量化，同时配套给出了各层次相关技术和与之匹配的算法，增强了可操作性。 在信息论的理论框架下给出了基于熵的多渠道信息冲突度的量化以及表述方法，借助于前人的理论和进一步的推证，构建出基于多层、多维理念的多渠道智能数据处理平台框架。 着重研究分析了目前求解大面积海域SST(海表面温度)的两大经典技术，即数据同化技术和卫星遥感反演技术，对这两中技术的优缺点进行了详细分析，尤其对其协同与互补性进行了对照，并给出了二维信息的融合方程。 在此基础上进行的算法研究，给出了基于证据组合规则的融合计算方法，尝试用不确知性信息融合技术解决SST问题。与数据同化等单一技术相比，经过改进的D-S组合规则可以较好地的发现和处理不确知性，适合SST问题特点；可能性计算填补了Bayes理论先／后验概率法则的盲区；模糊技术的应用可以对差值分级估计，便于融入经验数据。该方法能够有效地汇总多渠道信息，通过一致性和冲突评价实现多渠道优势互补。从表达方式上，软融合用估计误差分级和各级别可能性量化，乃至总体置信度取代简单的误差范围表示法，可以更确切地更多还原

【Abstract】 The Digital Ocean system needs various kinds of rich marine data and information that is densely distributed in space and continuous in time series. However, due to the vastness of ocean itself and the difficulty of obtaining information, it is impossible for the human being to get the information of an arbitrary point in ocean. Moreover, limitations of current marine observing methods result in the incertitude and even the conflicts among the information from multi-channel, which is due to many factors, such as error of measurement, deviation of computation of indirect information and so forth. Therefore, the fusion of multi-sourced data, especially of multi-channel incertitude information will not be avoided to establishing a Digital Ocean system. The key study of this dissertation is to find a new strategy to resolve this problem.Study on the applicational method of information fusion is still at the starting stage though scholars in ocean domain have already begun to probe into it in the ocean-related issues and a lot of documentation has already offered some successful application examples. Their study is limited in especial algorithm and application. The systematic study is very few. Tracing back its original reasons, they are that the professionalism of oceanography is an obstacle for the scholars in information fusion and that oceanographers’ weakness in information science and information fusion technology also holds back the popularization of this technology in the ocean-related domain.In order to solve the fusion of multi-channel incertitude information in Digital Ocean, more systematic study is carried on in the dissertation and the following results are achieved.The dissertation provides the layers and the way of information fusion for the first time, aiming at the Digital Ocean. Four-layer fusion model is proposed: data layer—use layer—service layer—advisor layer. Quantization is conducted to the granularity, character and method of information fusion in the Digital Ocean system.更多还原