Exploiting segmentation for robust 3D object matching

While Iterative Closest Point (ICP) algorithms have been successful at aligning 3D point clouds, they do not take into account constraints arising from sensor viewpoints. More recent beam-based models take into account sensor noise and viewpoint, but problems still remain. In particular, good optimization strategies are still lacking for the beam-based model. In situations of occlusion and clutter, both beam-based and ICP approaches can fail to find good solutions. In this paper, we present both an optimization method for beambased models and a novel framework for modeling observation dependencies in beam-based models using over-segmentations. This technique enables reasoning about object extents and works well in heavy clutter. We also make available a ground-truth 3D dataset for testing algorithms in this area.

Authors

• Michael Krainin
• Kurt Konolige
• Dieter Fox

Keywords

• Optimization
• Iterative closest point algorithm
• Estimation
• Clutter
• Computational modeling
• Data models
• Robot sensing systems
• 3D Point
• Iterative Closest Point
• Large Surface
• Evaluation Of Function
• Probabilistic Model
• Local Optimum
• Nonlinear Programming
• Segmentation Model
• Error Function
• Nonlinear Least Squares
• Feature Points
• Large Segments
• Pose Estimation
• Sensor Model
• Graphics Card
• Error Metrics
• Response Map
• Negative Log-likelihood
• Correct Matches
• Presence Of Occlusion
• Beam Model
• Correct Pose
• Color Information