MaLeStV

Project Details

Funding

FFG - KIRAS

Grant Number

873495

Duration

2019/09/01 - 2021/08/31

Contact

Martin Kampel

Persons

Martin Kampel
Thomas Heitzinger

Machine Learning of motion patterns in the penal system

The everyday life of a prison is characterised by aspects of security. A number of structural and organisational measures (prison cell design, video surveillance, …) are designed to reduce the risk of safety-related incidents in prisons and to support the personnel of the correctional system in their work. However, a permanent observation of potential endangering and endangered persons is not feasible due to limited personnel resources.

Violence in prison is no exception: suicides, for example, can be prepared and carried out in unobserved moments. Suicides in prisons often cause collateral damage beyond the tragedy of the event, ranging from increased media coverage to irritation among employees and inmates.

The project will provide important insights for the digitization of the prison regime system and offers the chance to develop highly innovative solutions from Austria for prisons worldwide.

The goals of this project are:

  1. The definition of behavioral patterns and scenarios in prisons to be detected by a 3D sensor
  2. Using 3D image analysis to register behavioral patterns of people in real-time to identify critical movement patterns
  3. Analyzing possibilities for the support of security personnel in prisons by detecting motion patterns
  4. Development of a prototype

Publications

T. Heitzinger and M. Kampel “A Foundation for 3D Human Behavior Detection in Privacy-Sensitive Domains”,
accepted at the 32nd British Machine Vision Conference (BMVC), 2021

S. Noiret, J. Lumetzberger, M. Kampel. “Bias and Fairness in Computer Vision Applications of the Criminal Justice System”, accepted at the IEEE Symposium Series on Computational Intelligence, 2021

T. Heitzinger and M. Kampel “IPT: A Dataset for Identity Preserved Tracking in Closed Domains”, in Proceedings of the 25th International Conference on Pattern Recognition (ICPR), Milan, Italy, January 2021

C. Pramerdorfer, M. Kampel, R. Kreissl “Behavior Detection as a Privacy-Enhancing Security Technology in Prison Cells”, Proc. 9th International Conference on Imaging for Crime Detection and Prevention (ICDP), London, UK, December 2019

C. Pramerdorfer, M. Kampel, M. Van Loock Multi-View Classification and 3D Bounding Box Regression Networks“, 24th International Conference on Pattern Recognition (ICPR), pp. 734-739, Beijing, China, August 2018

Project Partners

Das Projekt wird innerhalb des Sicherheitsforschungs-Förderprogramm KIRAS durch das Bundesministeriums für Verkehr, Innovation und Technologie (BMVIT) gefördert.