Evaluation method for automated BIM-capable object detection in tunnel constructions

Overview

Tunnel inspection and monitoring are essential for providing safe mobility in urban areas and transportation infrastructures. Project ABOUT is aimed to develop an advanced vision-based system that brings machine vision technologies and AI algorithms together to acquire high-resolution images from tunnel surfaces, automatically and efficiently. The data will be further processed to generate the 3D models of tunnel surfaces. In addition, a state-of-the-art Deep Learning algorithm is employed for damage detection as well as object recognition from tunnel images.

Research question

The proposed tunnel inspection system consists of different hardware and software components which should be selected and assembled based on the project requirements. The key-questions, which are aimed to be addressed in this project, are as follows.

  • Which type of the camera and how many of them should be used for high-speed image capturing in the tunnels with low illumination conditions?
  • Which type of LED should be used for providing sufficient light for the image acquisition system?
  • How could different measuring hardware be time-synchronized to get accurate results?
  • What is an optimum solution to process the big collected data from a long tunnel?

Procedure

The main components of the proposed system are machine vision cameras and light sources (e.g. Flash LEDs) as well as control units. All sub-systems are installed on a normal car and synchronized using a time synchronization unit. The intended operating speed is about 60-65 km/h which is suitable for traffic flows and high-speed monitoring with a minimum motion blur in the final images. The recorded images are processed in photogrammetry software like Agisoft Metashape or Pix4Dmapper to generate point clouds and 3D meshes. In addition, two different training datasets are manually generated for damage detection and object recognition tasks. The captured tunnel images are fed into a CNN (e.g. Deeplab V3+) which is well trained using the generated training datasets to detect different types of damages like cracks, spalling, rust, as well as tunnel objects such as signs, lights, cables, and so on.

 

Targeted results

There are three main desired results in this project: 3D models, damage maps, and object classification maps of tunnel surfaces which can be integrated into a Building Information Modeling (BIM) system. The project should strive for a relative accuracy of about 2 cm for 3D models and a surface resolution of about 3-5 mm for images.

ManagementProf. Dr. Gerrit Austen, Prof. Dr. Michael Hahn (Deputy)
PartnerViscan Solutions GmbH
Project e-mail addressGerrit.Austen@hft-stuttgart.de
FundingFederal Ministry for Economic Affairs and Energy (BMWi)
Call for proposalCentral Innovation Programme for Medium-Sized Enterprises (ZIM) - Cooperation Project
Duration06.05.2019 – 30.04.2021 extended until 31.07.2021

Team

Name and position Field Email and phone Room
Professor Geomatics +49 711 8926 2348 2/149
+49 711 8926 2560 2/209
Research Assistant Centre for Geodesy and Geoinformatics (ZGG)2/246