Abstract.
This paper proposes a novel method to exploit model similarity in  model-based 3D object recognition. The scenario consists of a large 3D model  database of vehicles, and rapid indexing and matching needs to be done without  sequential model alignment. In this scenario, the competition amongst shape  features from similar models may pose serious challenge to recognition. To  solve the problem, we propose to use a metric to quantitatively measure model  similarities. For each model, we use similarity measures to define a model- centric class (MCC), which contains a group of similar models and the pose  transformations between the model and its class members. Similarity informa- tion embedded in a MCC is used to boost matching hypotheses generation so  that the correct model gains more opportunities to be hypothesized and identi- fied. The algorithm is implemented and extensively tested on 1100 real  LADAR scans of vehicles with a model database containing over 360 models.