Streaming Monte Carlo Pose Estimation for Autonomous Object Modeling

Abstract:

This work contributes the optimization of a streaming pose estimation particle filter and its integration into an autonomous object modeling approach. The particle filter is advanced by an additional pose optimization in the particle weighting step. By integrating the method into the autonomous object modeling approach, the repositioning of objects is enabled, which is often necessary in order to acquire complete models. Experiments show that the usage of iterative closest point is too restrictive for general transformations. The used Monte Carlo method enables a robust pose estimation without loss of time and with high precision. Further, it is shown that the overall modeling results are improved clearly.

File Attach: 

Rink2016.pdf