Train operating companies are more and more interested in knowing the exact occupation state of their trains. Therefore, to detect the occupancy of each seat, new trains are equipped with seat occupancy sensors. These sensors are either installed in the seats (pressure sensors), or on the overhead luggage rack (radar, ultrasound, IR-sensors, optical...). However, all these solutions have a complex installation process (e.g. wire installation for the sensors and cut-outs into the luggage rack) and limit themselves to seat occupancy.
The goal of this thesis is to research the feasibility of combining regular 2D camera images with other sensors (e.g. LIDAR). In the thesis the student will need to develop new systems to combine the data and allow models to reason on such fused data.
The thesis should include the following phases:
Requirement definition based on Televic products (power, budget, ...)
Creating a measurement setup for data aggregation
Developing the data fusion algorithms
Training machine learning models on top of a dataset existing of data from different sensors.
The student will have the opportunity to go through the complete development phase of a machine learning project with state of the art sensors.
Nature of the work
Specialty: AI / Machine Learning
Type of work: Research: 40%, Implem.: 30%, Experim.: 30%
Location: Televic, University
Type of activities: Experimenting, Implementation, Literature study, Programming