Project: Instantaneous Visual Inspection of High-resolution Engineering Construction Scans
To guarantee for the safety of usage, large buildings and infrastructure such as bridges need to be checked on a regular basis. Cracks in the constructions must be detected, measured, and tracked over time. Based on the observations, decisions must be taken concerning the COtenance measures. Correct decisions are of utmost importance to not risk any accidents and casualties. The topic is of particular interest to a country like Switzerland, where bridges are widely been used for road construction in a mountainous landscape. Modern technology nowadays supports these inspection and decision-making tasks in form of scanning the bridges with high-resolution 3D scanners and taking pictures with high-resolution cameras. The ever increasing information that becomes available is a challenge in terms of handling the large amount of data, but also allows for the deployment of new processing methods._x000D__x000D_Traditionally, the scans are handled by reconstructing the geometry of the scanned surfaces in a 3D environment using triangular meshes or fitting planes/surfaces, while the pictures are being looked at separately. This reconstruction step scales very poorly with increasing complexity of the gathered data. As a consequence, it is impossible to immediately check the scanning results on site. We propose to avoid such a reconstruction step by directly operating on the scanned data. The scanned data is given in form of so-called point clouds, i.e., points on the 3D surfaces without any connectivity or neighborhood information. The rendering of such point clouds have become state of the art in computer graphics in the recent two decades. However, pure image-space point cloud rendering techniques are the only ones that can produce proper renderings without requiring any neighborhood information. As a consequence, they scale well to the amount of data one has to handle when scanning bridges with high-resolution 3D scanners. On-site investigation of the scanning results will become possible. The resulting software will be made available to the market in form of a service package by the enertec engineering ag._x000D__x000D_Technical challenges that we will be facing in this project are (1) multiresolution representations for image-space point cloud rendering to deal with data sizes that cannot be loaded to the CO memory of a standard computer ("out-of-core processing"), (2) on-the-fly image-space registration of scans from multiple scanner positions, which is necessary to adjust to different lighting conditions and will, again, be done without any pre-processing of the data, (3) integration of high-resolution photo material in 3D point cloud representation to allow for full-detail crack investigation within the 3D rendering environment, (4) integration of meta information, like supplemental documents, videos, or 3D models, in the 3D environment with proper location mapping, and (5) visual investigation of changes in cracks over time using coordinated views of derived or measured attribute spaces at respective locations._x000D__x000D_The consortium consists of one SME P (enertec engineering ag, Winterthur, Switzerland – project leader) and two university Ps (Jacobs University, Bremen, and Chemnitz University of Technology, both Germany). Enertec engineering ag is an SME that provides, among others, services for bridge inspection. Services include providing equipment and expertise for bridge inspection to civil engineering offices and executing the entire inspection procedure for public clients. The visualization and computer graphics group at Jacobs University, Bremen, which is led by Prof. Linsen, has significantly pushed the state of the art in image-space point cloud rendering. To be mentioned is the "Points InSight" software that came out of these research efforts. A CO contributor to these developments was Jun.-Prof. Rosenthal, who is now leading the visual computing group at the Chemnitz University of Technology. Also, both investigators have jointly developed novel approaches for coordinated views of spatial and attribute space views, including their contribution that won the IEEE Visualization 2008 Design Contest.
|
Acronym
|
enercloud
(Reference Number: 7001)
|
|
Duration
|
01/04/2012 - 10/09/2015
|
|
Project Topic
|
The condition of buildings and infrastructure is regularly checked using high-resolution 3D scanners and cameras. We propose a novel approach to handle the huge amount of data using an image-space point cloud representation that allows for immediate interactive inspection of the gathered data.
|
|
Network
|
Eurostars
|
|
Call
|
Eurostars Cut-Off 7
|
Project partner
