Radar Multi-target Track Correlation Based On FCM

In recent years, fuzzy cluster analysis was widely used in many fields and has achieved satisfactory project effects. Its application area includes pattern recognition, image processing, water quality analysis, etc.Data association is among the key technology problems which need to be solved when radar is in the process of multi-target tracking filter. The accuracy of traditional track association will lower under the circumstance of complex and multi goals' environment as well as cross track. This study will present a specific discussion on the application of fuzzy cluster analysis over radar multi-target track association. This study applied fuzzy C-means clustering for data association in order to implement the data association and accurate tracking towards multi-target. Moreover, simulation experiment results were provided for the support of effectiveness of this method. This thesis, depriving from the basic Sci & Tech researching project of "the research of digital processing method of intermediate frequency signal for marine radar " supported by the Ministry of Communication, is mainly researching on the processing method of. radar target tracking.This study can be divided into the following parts: Firstly, basic concepts and theories about data association, related gate set up, radar plot and track association procedure, the Kalman Filter Algorithm, etc, were introduced. This study also discussed several typical data association algorithm, cross-tracking, Nearest-Neighbor Data Association Algorithm, Probabilistic Data Association Algorithm, Joint Probabilistic Data Association Algorithm, etc. Applicable environment and performance would be analyzed simultaneously.Secondly, this study introduced fundamental principle of fuzzy c-means(FCM): characteristics of fuzzy classification were introduced in target tracking system; in the case of known destination array, accumulate the degree of membership between measuring plot and target predicted position; then combined Kalman Filter and weight the predictive vector with the degree of membership as coefficient to update target state estimation. Simulation result demonstrates that under the circumstances of high-density clutter and complex and multi goals, fuzzy cluster algorithm processes superiority in some degree, that is, less amount of calculation, and guarantee of high degree of accuracy tracking towards multi-target.At last, two kinds of improvement were suggested. One is that making clutter as cluster which can fit reality better, as well as lower the incorrect association probability caused by clutter or interference plot. The other is to preserve the plots by cross-tracking when the degree of membership between two or more plots and one identical target are the same.