Within the scope of the project, a measuring and evaluation method is to be developed with which the surface of a tunnel can be digitized so precisely that the recorded data can be used to carry out deformation monitoring by surface and depth inspection. The aim is to achieve an accuracy of position and depth measurement of about 1 mm at a speed of up to 65 km/h.
In addition, all equipment features (e.g. signs, markings, lane signals, lighting equipment, hydrants, loudspeakers, etc.) are to be detected automatically using deep learning methods.
For the development of the planned method, solution approaches from photogrammetry / videogrammetry / computer vision as well as from deep learning are used. 3D point clouds can be generated by a suitable acquisition geometry with a multi-camera setup and the successive intelligent linking of overlapping image acquisitions. Point clouds and images are input for CNNs (e.g. Dynamic Graph CNNs), from which local geometric features are captured after successful training, a semantic segmentation is performed and objects are classified and recognized automatically. During object formation, care is taken to ensure that the object representation is BIM compatible.
|Management||Prof. Dr. Gerrit Austen, Prof. Dr. Michael Hahn (Deputy)|
|Partner||Viscan Solutions GmbH|
|Project e-mail address||Gerrit.Austen@hft-stuttgart.de|
|Call for tender||ZIM|
|Duration||01.01.2018 – 31.12.2022|
|Name and position||Field||Email and phone||Room|
|Fakultät Vermessung, Informatik und Mathematik||2/246|