双红外传感器分布式信息融合的研究

【作者】 陈非石；

【导师】 孔祥维；

【作者基本信息】 大连理工大学 ， 信号与信息处理， 2004， 硕士

【摘要】 随着多传感器技术的快速发展，传感器获取的信息的种类和数量大大增加，多传感器信息融合技术应运而生。针对雷达／红外、红外／紫外、紫外／紫外等多传感器信息融合的应用，世界各国的研究人员经过探索和研究，纷纷提出了一些重要的观点和算法，使多传感器信息融合技术迅速发展起来。 本文主要针对双波段红外传感器的特点进行了信息融合方面的研究。旨在解决双红外传感器信息融合技术的实际应用问题。主要研究内容包括分布式信息融合结构的改进，具体的航迹关联、未关联航迹的处理、航迹融合，目标威胁估计五个方面。 针对双波段红外传感器，本文研究并实现了统计和模糊航迹关联算法，包括：加权法、修正法、序贯法、双门限法、NN、KNN、MKNN、模糊双门限法。经过对算法的研究和仿真实验结果的分析，发现这些算法在目标密集、存在交叉的情况下关联性能会严重下降，导致较多的错漏关联。另外，基于统计的航迹关联算法对跟踪滤波器的类型存在限制，跟踪性能的改变直接影响航迹关联的正确率。为了克服以上问题，本文基于模糊集和序贯检测理论的思想，结合双波段被动式红外传感器的特点，提出了一种针对双波段被动式红外传感器的模糊序贯航迹关联算法，仿真实验证明该算法性能优于其他航迹关联算法，更具有实际适用性。 针对传统的“与／或”未关联航迹输出方法增加漏警虚警概率的问题，本文结合红外传感器特点，基于意见一致性决策理论，提出了对未关联航迹的一种新的处理方法。 航迹融合是多传感器信息融合中的一个重要环节，关系到信息融合结果的精确性。本文针对红外传感器的情况，实现和研究了加权、互协方差和自适应三种航迹融合方法。并进行了仿真实验分析和比较。针对系统的实时性要求，本文提出了自适应航迹融合算法的简化方法，并通过仿真实验证明了算法的有效性。 威胁估计是对目标威胁程度的估计。本文根据红外传感器的特点，提出了一种多属性决策的威胁估计的方法。更多还原

【Abstract】 With fast development of multi-sensor technology, sensors obtain more kinds and quantities information, multi-sensor information fusion technology plays an important role. Researchers from many countries propose important views and algorithms for applications of radar/infrared, infrared/UV, UV/UV multi-sensor fusion, multi-sensor fusion technology develops rapidly.This thesis researches information fusion methods according to characters of dual-band infrared sensors. This thesis brings forward an improved distributed information fusion structure for its economy and rational structure. And studies and improves track correlation, track fusion and threaten estimation algorithms.This thesis studies and realizes traditional track correlation methods: weighting method, WCF, double threshold, NN, KNN, MKNN etc according to infrared sensors characters. Through studying algorithms and analyzing simulation experiments results, this thesis finds that performance of these algorithms decline seriously when targets are dense and across. Also, track correlation algorithms based on statistical theory limit type of tracking filter, and change of tracking performance influence the correct rate of correlation directly. For avoiding these problems, based on fuzzy set and sequential detection theory, combining characters of dual-band infrared sensors, the thesis brings forward a fuzzy sequential track correlation algorithm, and proves that performance of the algorithm is better than others through simulation experiments.Traditional and/or methods for disposing uncorrelated tracks have influence on leak/false rate. The thesis brings forward a new method based on consensus theory, according to characters of infrared sensors.Track fusion is an important part of information fusion. It concerns accuracy of the result of information fusion. This thesis achieves and studies weight, WCF and adaptive track fusion algorithms, analyses and compares these algorithms through simulation experiments. In addition, because of time limit, the thesis brings forward a predigesting method of self-adaptive track fusion and proves it’ s effective by experiments.Threaten estimation is to analysis the degree of target threaten. Combining characters of infrared sensors, this thesis brings forward a threaten estimation method based on property decision.更多还原