Fuzzy Clustering For Data Fusion In A Recognized MaritimePicture

摘要

The generation and maintenance of a RecognizernMaritime Picture consists in part of associating contactrnreports from sensor sources with existing tracks orrninitiating new tracks. This decision-making process oftenrntakes place in an environment in which the goal and thernconstraints are not known precisely. To deal quantitativelyrnwith imprecision, we usually employ the concepts andrntechniques of probability theory. The use of a probabilisticrnapproach requires that the imprecision can be equated withrnrandomness and that the characteristics of this randomnessrnare reasonable well known. This is generally not the casernfor the generation of the RMP. An alternative approach isrnto view the problem as fuzzy decision making and to employrnthe concepts and techniques of fuzzy sets The approachrnproposed in this paper makes use of the non-real-timernnature of the problem to make maximum use of thernavailable data. The proposed approach makes use of thernreverse Cuthill-McKee ordering technique to establish anrninitial estimate of the number of clusters and the c-meanrnclustering techniques refine the cluster and to establishrnfuzzy membership functions. The effectiveness of thernapproach is demonstrated on data acquired off the EastrnCoast of Canada.