Recognition and Location by Parallel Pose Clustering

This paper describes the recognition and location of 3D object models from depth-based data using a parallel pose clustering algorithm. We describe a leader-based partitional clustering algorithm and demonstrate a successful parallel implementation of this. Results are presented for a variety of synthetic and real scene data. We also consider how the basic approach can be extended to recognise objects from a single 2D intensity image by perspective inversion. Our eventual aim is to combine such a dual data approach within a single parallel system.