A learning-based variable size part extraction architecture for 6D object pose recovery in depth images

作者：

Highlights：

• We propose part-based Iterative Hough Forests (IHF) for 6 DoF object registration.

• The parts are represented with scale-variant Histogram of Control Points features.

• Scale-variability of the features is iteratively exploited for pose refinement.

• The refinement is improved by an automatic variable size part extraction framework.

摘要

•We propose part-based Iterative Hough Forests (IHF) for 6 DoF object registration.•The parts are represented with scale-variant Histogram of Control Points features.•Scale-variability of the features is iteratively exploited for pose refinement.•The refinement is improved by an automatic variable size part extraction framework.

论文关键词：Object registration,6 DoF pose estimation,Scale-variant HoCP feature,One class training,Random forest,Iterative refinement

论文评审过程：Received 2 September 2016, Revised 27 March 2017, Accepted 13 May 2017, Available online 21 May 2017, Version of Record 30 May 2017.

论文官网地址：https://doi.org/10.1016/j.imavis.2017.05.005