UAV ground target detection and tracking

Connected publications:

  • Máttyus Gellért, Benedek Csaba, Szirányi Tamás: Multi Target Tracking On Aerial Videos. In Proc. of ISPRS Workshop 2010: Modeling of optical airborne and space borne sensors, 2010.
  • L. Kovács, Cs. Benedek: Visual Real-time Detection, Recognition And Tracking Of Ground And Airborne Targets. In Proc. of SPIE-IS&T Electronic Imaging - Computational Imaging IX, SPIE vol. 7873, pp. 1-12, 2011.
  • G. Hummel, L. Kovács, P. Stuetz, T. Szirányi: Data Simulation And Testing Of Visual Algorithms In Synthetic Environments For Security Sensor Networks. In Proc. of Future Security 2012 - Security Research Conference, Communications in Computer and Information Science (CCIS), Spinger, vol. 318, pp. 200-203, 2012.

 

 

Data Simulation and Testing of Visual Algorithms in Synthetic Environments for Security Sensor Networks







To appear in proceedings of Future Security - Security Research Conference, 2012.

Georg Hummel1, Levente Kovács2, Peter Stütz1, and Tamás Szirányi2

1Institute of Flight Systems, Universität der Bundeswehr München, Neubiberg, Germany
http://www.unibw.de/lrt13
2Computer and Automation Research Institute, Hungarian Academy of Sciences, Hungary
Distributed Events Analysis Research Laboratory
http://web.eee.sztaki.hu

Abstract. Current development of security sensor networks and their processing algorithms use pre-recorded or abstract data streams for testing, often missing important ground truth for validation. This paper proposes a simulation-based test bed, presenting an approach to use commercial off the shelf virtual reality environments to create adaptive simulations of visual and non visual sensors, scenarios and data streams for testing of perceptive algorithms. An example using airborne visual multi object tracking is discussed and validated.
Keywords: virtual reality, security networks, computer vision, target tracking

fig1
fig2
fig3
Real samples
Sim samples

The presented test approach demonstrated the viability of using simulated data for testing and development of perceptive modules (especially visual algorithms), and shows general usability for detection and tracking. The lack of optical and signal errors simplifies processing in synthetic feeds, but these issues will be addressed in future work. The benefits of such an approach are the availability of ground truth data with automatic evaluation, closed loop simulations and prompt repeatability. Also, testing classification and identification methods is of interest, due to the limited variety in terms of textures, and entities.


© 2012 by Institute of Flight Systems, Universität der Bundeswehr München and Distributed Events Analysis Research Laboratory at the Computer and Automation Research Institute, Hungarian Academy of Sciences.