REGISTERING INCOMPLETE RADAR IMAGES USING THE EM ALGORITHM

This paper describes an application of the EM (expectation and maximis ation) algorithm to the registration of incomplete millimetric radar i mages. The data used in this study consists of a series of non-overlap ping radar sweeps. Our registration process aims to recover transforma tion parameters between the radar-data and a digital map. The tokens u sed in the matching process are fragmented line-segments extracted fro m the radar images which predominantly correspond to hedge-rows in the cartographic data. The EM technique models data uncertainty using Gau ssian mixtures defined over the positions and orientations of the line s. The resulting weighted least-squares parameter estimation problem i s solved using the Levenberg-Marquardt method. A sensitivity analysis reveals that the data-likelihood function is unimodal in the translati on and scale parameters. In fact, the algorithm is only potentially se nsitive to the choice of initial rotation parameter; this is attributa ble to local sub-optima in the log-likelihood function associated with pi/2 orientation ambiguities in the map. By adopting Levenberg-Marqua rdt optimisation we reduce the local convergence difficulties posed by these local rotation maxima. The method is also demonstrated to be re latively insensitive to random measurement errors on the line-segments . (C) 1997 Elsevier Science B.V.