Leveraging local and global descriptors in parallel to search correspondences for visual localization

作者：

Highlights：

• A visual localization method that leverages local and global descriptors in parallel.

• A weighted hamming regularization term for training local descriptors.

• A new probability model for priority search strategy in random trees.

• Extensive experiments show that the proposed localization method is very effective.

摘要

•A visual localization method that leverages local and global descriptors in parallel.•A weighted hamming regularization term for training local descriptors.•A new probability model for priority search strategy in random trees.•Extensive experiments show that the proposed localization method is very effective.

论文关键词：Visual localization,6DoF pose,Parallel search,Learning based descriptor

论文评审过程：Received 13 August 2020, Revised 28 August 2021, Accepted 20 September 2021, Available online 22 September 2021, Version of Record 1 October 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108344