多传感器融合数据信息的定量度量方法研究

With the development of science and technology, the conditions of using multi-sensor system to acquire real world information more accurately and reliably are becoming mature gradually, and the application demands to it are increasing. In recent years, although many data fusion methods are proposed and developed, there is a crucial and primary issue that is not well investigated; it is how to define the correlation extent of different information in multi-variable system in theory, and how to measure and evaluate the information and fusion performance quantificationally. In this project, our research is concentrated on the following three aspects of this problem. (1)Theory and method innovation: In order to measure the correlation degree of signal from multi source, a new concept called correlation information entropy is put forward based on the concepts in classical information theory; the corresponding approach of information description and the calculating formula are established to measure the correlation degree and overlap extent of each variable in multi-variable system quantificationally. (2)The extend of fusion and measure method: Combined with correlation information entropy theory, rough set theory and fuzzy measure are used in information fusion to resolve the problems such as information reduction, rules extraction and optimizing, similarity and similarity measure, etc. (3)Theory and practice incorporation: These methods and theory proposed are applied to and approved by hyper-spectral data processing and complex control systems fault diagnosis. Experiments results show that the new methods are effective in data classification and compressing, extraction of fault characters and evaluation of fusion methods.

研究多传感器数据信息的定量描述方法及相关技术，为评价和支持多传感器配置和信息融合算法的实现及应用奠定基础。拟综合采用相关理论及信息熵理论，创造性地提出相关信息熵概念，统一评估传感器配置和信息融合算法的准则指标。该研究不仅可促进多传感器信息融合理论与应用的发展，而且对信息处理领域中的信息定量描述提供一种新的思路和途径。