Analysis and processing of mobile mapping data (LIDAR, RGB) in railroad environment

ACQUAL

Potential supervisors

Ville Lehtola

Spatial Engineering

This topic is not adaptable to Spatial Engineering

Suggested Electives

Laser scanning 201800310, Positioning and Imaging Technology 201700167, Programming courses (c++/python), Courses dealing with machine/deep learning, courses about photogrammetry and/or computer vision

Additional Remarks

Internship, no Fieldwork. The topic may include a machine learning context, depending on the student. Good programming skills are mandatory. Knowledge of deep learning software libraries is a plus (e.g., Tensorflow).

Description

Asset management is important for railroad management. Information about the assets is obtained from the data captured from a mobile mapping system attached onto a train. However, the analysis and processing of this LIDAR and RGB data is not trivial. There are multiple problems related to sensor fusion, object detection, object classification, and object localization. The data set will be available for the student in August for the detailed development of the proposal.

This research is done in collaboration with a big Dutch company, Fugro. The research includes an internship at the company. The student must have good skills in computational science (c++/python) and in communication to satisfy the academic and internship requirements. Mobile mapping data includes laser scanned point clouds and RGB imagery that are both geo-referenced.

Objectives and Methodology

Machine/deep learning approaches for sensor fusion, object detection, object classification, and object localization methods applied on RGB images and 3D point clouds. Reporting to both the university (MSc research) and the company (internship).

Further reading