RECOGNITION OF PARAMETERIZED OBJECTS FROM 3D DATA - A PARALLEL IMPLEMENTATION

We present a parallel implementation of an algorithm designed to recognize fully parameterized objects from 3D data. For the purpose of recognition, the mapping of sensed features with model features is structured as an interpretation tree. However, with the use of parameterized models, most of the computation performed during the tree search is actually spent in the updating step of the object parameters, this process being executed through a SUP/INF network. A major contribution of our implementation is the exploitation of the intrinsic parallelism of the SUP/INF network, while the interpretation tree search is also distributed on our transputer based MIMD architecture.