Object Recognition in Point Cloud Data.

Kurzbeschreibung / Ausgangslage / Project Description.

Information regarding the surrounding areas around tracks is collected by SBB using a dedicated system mounted on a diagnostic vehicle. Sections of 14000 points are collected at each 68 mm along the track. This data is to be used in determining distances to closest objects and assess if safety spaces are observed along the track.

Environmental conditions affect the collection of data and produces artefacts that need to be filtered out. Currently SBB produces targeted analyses of the data only for limited area of interest, analyses that are largely manual. The information collected can be further used to build an inventory of infrastructure assets if assets can be identified in an automated process.

Aufgabenstellung / Ziele / Objectives.

Develop algorithms and/or models to identify artifacts and assets in the delivered data. Concretely, the project includes the use of (un)supervised machine learning libraries, e.g. TensorFlow 2, to develop assets identification methods.

Anforderungen /Requirements (Studienrichtung, Studiendauer).

Master student in Engineering, Machine Learning, Statistics, Mathematics, Physics

Machine Learning and/or Software development background.

Fachgebiet / Themen.

Ingenieurwesen / Technik

Ort.

Bern

Zeitraum.

4-6 Months

Sprachkenntnisse / Languages.

Deutsch, English

Kontaktadresse / Contact.

Lucian Ancu (I-VU-UEW-MUD-TEN), lucian-stefan.ancu@sbb.chLink öffnet in neuem Fenster.

Form der Bewerbung und gewünschte Dokumente / Application Documents.

Lebenslauf / Curriculum Vitae

Weiterführender Inhalt