Classification of radar signatures by autoregressive model fitting and cluster analysis

Ninety radar signatures were digitally recorded during the probing of ten different land and marine objects with a pulsed coherent Doppler radar. Their spectra were evaluated on the basis of the Marple algorithm for autoregressive model fitting. An attempt at classification is made. The method consists in representing the radar signatures as points in four-dimensional space and identifying the obtained clusters of four-dimensional points with the observed objects. The cluster analysis is carried out assuming that the classification parameters have different clustering lengths