UAS Image-Based Point Clouds to 3D BrIM: Deep Semantic Segmentation

PI: Yelda Turkan (OSU), yelda.turkan@oregonstate.edu, ORCID: 0000-0002-3224-5462

Co PIs: none

AMOUNT & MATCH: $40,000 from PacTrans; $40,000 Match

PERFORMANCE PERIOD: 8/16/2019 – 8/15/2021

STATUS: Completed

CATEGORIES: Point Clouds, Bridge Inspection, UAS

RESEARCH PROJECT HOT SHEET:

UTC PROJECT DOCUMENTATION:

FINAL PROJECT REPORT:

PROJECT DATA:

DESCRIPTION: This project will develop a novel framework that automatically converts 3D dense point clouds, obtained from 2D digital images collected using an UAS, into 3D BrIM.

The framework will make it much more convenient and faster to implement 3D BrIM, thus improve the current bridge inspection and management practice in terms of efficiency and safety. Computer vision algorithms will be used to segment and label bridge components automatically using deep learning techniques.

The anticipated outcomes of this research will be to a) provide a trained model that enables to automatically identify and label bridge elements from their 3D point clouds; b) speed up the transition from 2D to 3D format in bridge inspection and management practices; c) increase the adoption of 3D BrIM for maintenance phase of infrastructure projects. All of these factors should help assist in maintaining U.S. bridges in a state of good repair, thus help ensure public mobility.

DELIVERABLE DUE DATE DATE RECEIVED
Research Project Progress Report #1 4/10/2020 4/2/2020
Research Project Progress Report #2 10/10/2020 10/12/2020
Research Project Progress Report #3 4/10/2021 4/1/2021
Research Project Progress Report #4 10/10/2021 10/13/2021
No Cost Extension Request 6/15/2021 6/6/2021
Draft Report 12/15/2021 12/15/2021
Final Project Report 2/15/2021 5/10/2021