Abstract
We propose a novel approach for verifying model hypothe- ses in cluttered and heavily occluded 3D scenes. Instead of verifying one hypothesis at a time, as done by most state-of-the-art 3D ob ject recog- nition methods, we determine ob ject and pose instances according to a global optimization stage based on a cost function which encompasses geometrical cues. Peculiar to our approach is the inherent ability to de- tect significantly occluded ob jects without increasing the amount of false positives, so that the operating point of the ob ject recognition algorithm can nicely move toward a higher recall without sacrificing precision. Our approach outperforms state-of-the-art on a challenging dataset including 35 household models obtained with the Kinect sensor, as well as on the standard 3D ob ject recognition benchmark dataset.