[Automatic Systems 1] Person detection using time-of-flight camera and machine learning | BEAMS

[Automatic Systems 1] Person detection using time-of-flight camera and machine learning

Project information
Project type: 
Master thesis
Academic year: 
2017-2018
Status: 
Proposal
Research unit: 
Embedded electronics
BEAMS supervisors
Supervisor
External supervisors
Kevin De Cuyper
Industrial promoter

Automatic Systems has developed an extensive portfolio of turnstiles as shown on AS website. One of the main issues in the high-security market is to accurately detect and count persons in the turnstile. The detection method must be resilient to a wide range of lighting conditions, person morphologies, person behaviors, frauds,… Automatic Systems is developing a new detection method using time-of-flight camera (see IFM website for details on the sensor). The detection software that has been developed is based on computer vision methods.

The aim of the proposed thesis is to apply machine learning methods to the person detection problem. These methods should replace or extend the algorithms developed by Automatic Systems until now. An additional problem that one should take into account is that the person detection executes in quasi-real time and that the SW must run on a low-cost, embedded computer.

The main work packages are:

  1. Review and if required, extend the available database of 3D images and video sequences,
  2. Select a subset of machine learning methods (neural networks, support vector machine,…)  that could be applied,
  3. Develop the selected algorithms and evaluate their performances,
  4. Investigate the SW usability on an embedded computer

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer