Assessing losses for point set registration
2020 (English)In: IEEE Robotics and Automation Letters, E-ISSN 2377-3766, Vol. 5, no 2, p. 3360-3367, article id 9013051Article in journal (Refereed) Published
Abstract [en]
This letter introduces a framework for evaluation of the losses used in point set registration. In order for a loss to be useful with a local optimizer, such as e.g. Levenberg-Marquardt, or expectation maximization (EM), it must be monotonic with respect to the sought transformation. This motivates us to introduce monotonicity violation probability (MVP) curves, and use these to assess monotonicity empirically for many different local distances, such as point-to-point, point-to-plane, and plane-to-plane. We also introduce a local shape-to-shape distance, based on the Wasserstein distance of the local normal distributions. Evaluation is done on a comprehensive benchmark of terrestrial lidar scans from two publicly available datasets. It demonstrates that matching robustness can be improved significantly, by using kernel versions of local distances together with inverse density based sample weighting.
Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers Inc. , 2020. Vol. 5, no 2, p. 3360-3367, article id 9013051
Keywords [en]
Performance evaluation and benchmarking, probability and statistical methods, Geometry, Inverse problems, Maximum principle, Normal distribution, Expectation Maximization, Levenberg-Marquardt, Local optimizers, Point-set registrations, Terrestrial lidars, Violation probability, Wasserstein distance, Benchmarking
National Category
Natural Sciences
Identifiers
URN: urn:nbn:se:ri:diva-45287DOI: 10.1109/LRA.2020.2976307Scopus ID: 2-s2.0-85081718613OAI: oai:DiVA.org:ri-45287DiVA, id: diva2:1454231
Note
Funding details: VINNOVA; Funding text 1: Manuscript received September 10, 2019; accepted February 3, 2020. Date of publication February 26, 2020; date of current version March 9, 2020. This letter was recommended for publication by Associate Editor L. Paull and Editor S. Behnke upon evaluation of the reviewers’ comments. This work was supported by ELLIIT (a Strategic Area for ICT research, funded by the Swedish Government), and in part by the Vinnova through the Visual Sweden network. (Corresponding author: Per-Erik Forssén.) Anderson C. M. Tavares is with the Computer Vision Lab, Department of Electrical Engineering (ISY), Linköping University SE-581 83, Linköping, Sweden, and also with RISE SICS East Linköping SE-581 83, Linköping, Sweden (e-mail: anderson.tavares@ri.se).
2020-07-152020-07-152024-01-17Bibliographically approved