SUPR
Object Detection in Lidar Point Clouds
Dnr:

NAISS 2024/22-239

Type:

NAISS Small Compute

Principal Investigator:

Per-Erik Forssén

Affiliation:

Linköpings universitet

Start Date:

2024-03-01

End Date:

2025-03-01

Primary Classification:

10207: Computer Vision and Robotics (Autonomous Systems)

Allocation

Abstract

The proposed project is a pilot study for a collaboration with Siemens Energy. Siemens Energy is developing digital twins of their environmentally friendly power plants (with solar cells, electrolysis and hydrogen turbines). A digital twin is an interactive and immersive simulation environment that accurately corresponds to a particular physical environment, and Siemens Energy use these for staff training and simulated operation. Current digital twins of power plants have been created by hand, using CGI-modelling, but to scale up the use of digital twins, the process must be simplified. This pilot study will therefore investigate the use of 3D-scanning in the generation of digital twins. Siemens will provide CVL with a 3D lidar scan of a power plant (a detailed point cloud), and with CAD-models of objects of interest at the plant. CVL will investigate state-of-the-art in lidar point-cloud based object detection, with the purpose of refining the point cloud 3D model into a CGI-model with annotated components. This will later allow the detected components to be replaced with existing simulations of engines and interface components, that allow metadata about components to called up by a user during XR interaction with the digital twin.