MULTIPLE LINE-TEMPLATE MATCHING WITH THE EM ALGORITHM

This paper shows how multiple shape hypotheses can be used to recognis e complex line patterns using the expectation-maximisation algorithm. The idea underpinning this work is to construct a mixture distribution for an observed configuration of line segments over a space of hypoth esised shape models. According to the EM framework each model is repre sented by a set of maximum likelihood registration parameters together with a set of matching probabilities. These two pieces of information are iteratively updated so as to maximise the expected data likelihoo d over the space of model-data associations. This architecture can be viewed as providing simultaneous shape registration and hypothesis ver ification. We illustrate the effectiveness of the recognition strategy by studying the registration of noisy radar data against a database o f alternative cartographic maps for different locations. (C) 1997 Else vier Science B.V.