Project: #192 Low-Cost 3D model acquisition for rapid accident investigation Progress Report - Reporting Period Ending: Sept. 30, 2018 Principal Investigator: Christoph Mertz Status: Active Start Date: Aug. 1, 2017 End Date: June 30, 2019 Research Type: Applied Grant Type: Research Grant Program: FAST Act - Mobility National (2016 - 2022) Grant Cycle: 2017 Mobility21 UTC Progress Report (Last Updated: Sept. 25, 2018, 3:38 p.m.) % Project Completed to Date: 50 % Grant Award Expended: 56 % Match Expended & Document: 61 USDOT Requirements Accomplishments We have been working on two different kinds of tools for accident reconstruction. The first completes partial point clouds and the second fits primitives to point clouds. When crashes are modeled by reconstructing the scene from images there are often holes and missing areas in the model. They correspond to parts that have little or no texture, are reflective, or transparent. Simple hole filling algorithms make the assumption that holes can be filled with a plane, without considering the context of the hole. The shape completion tool is a learning method, where the system learns the possible shapes from examples and learns how to best complete a model. Fitting primitives will help in analyzing the crashed vehicle. We anticipate that it will be able to identify the various parts of the vehicle, even if they are disbursed or deformed. Impacts The tools developed in this project will help in automatically analyzing vehicle crashes. Other Supplemental material to the publication has been uploaded. Updates to the project are posted to the project web site http://www.cs.cmu.edu/~reconstruction/ Outcomes New Partners nothing to report Issues We tried to visit a facility that does car crashes for the automotive industry, but our contacts have so far not been able to find a customer who would be willing to have us as an observer.