Surveillance systems Annual report 2012

Authors:

  • Stefan Nilsson
  • Staffan Abrahamson
  • Maria Andersson
  • Erika Bilock
  • Magnus Gustavsson
  • Hans Habberstad
  • Gustaf Hendeby
  • Mikael Karlsson
  • Håkan Larsson
  • Dietmar Letalick
  • David Lindgren
  • Staffan Lindström
  • Fredrik Näsström
  • Jörgen Karlholm
  • Joakim Rydell

Publish date: 2012-12-28

Report number: FOI-R--3562--SE

Pages: 33

Written in: Swedish

Keywords:

  • sensor systems
  • sensor networks
  • detection
  • target tracking
  • classification
  • camp
  • protection
  • multi sensor fusion
  • urban environment
  • urban operations
  • anomaly detection
  • deviation detection
  • surveillance
  • blue force tracking
  • SLAM

Abstract

This annual report gives an account of the activities carried out and the results in the first year of the three-year Armed Forces' project Surveillance Systems. The project studies how various sensor systems can co-operate to accomplish improved situation awareness in the urban environment. With multi-sensor data as a base, important research tasks are here to devise robust data fusion methods and algorithms for detection, classification, and tracking of human movements and to develop methods for automated detection of deviations from the normal state. The research efforts of the project are concentrated on the following military need areas: Continuous Surveillance, Crowd Surveillance, and Blue Force Tracking. The project participates in several international co-operations, viz., three EU FP7 projects and the NATO group Multi- Sensor Integration in Urban Operations. A special focus of the project is to study the design of multi sensor systems for detection, tracking and classification of threats against military camps. A completed need analysis with military expertise has identified major threat indicators. How to detect these threat indicators with different sensor configurations is studied in MSSLab. The detection work has been focused on the development of methods for automatic discovery of anomalous behaviour. For more robust tracking a full body detector has been combined with a "head"-detector, and the result produces a richer description of the movements of persons in a scene. A field experiment has been carried out with optical and acoustic sensors recording vehicles and a human in motion. The data set will be used for system related analysis where especially fusion of acoustics and optics is studied. In a special work the properties of radar as supporting sensor in a multi sensor system have been investigated. Among other things, it is found that the all-weather capacity and the capability of detecting Doppler are important complements to optical sensors. The automated method for determination of normal state and for anomaly detection of a crowd has been developed further. Especially, we have tested to use detections of persons as input data to a new function that continually looks for groupings. In this way one gets an idea at every instant whether persons are moving individually or in groups (or if groups are being formed or split). We have then used the change in the spatial extension of the group, from one instant to another, as input data to the anomaly detection in order to simultaneously estimate the activity of the group. Tests with annotated data (from recorded scenarios) display preliminarily a prediction capability for the new function of more than 80 %. Previously, we have used a more coarse detection method (based on optical flow), which cannot be used in the same way to analyze groups. In the area Blue Force Tracking new techniques are investigated to follow one's own soldiers in buildings where GPS coverage is missing. With our own-developed test system CHAMELEON - consisting of a stereo camera, an IMU (Inertial Measurement Unit) and now also a laser pointer - we have by using the SLAM-technique (Simultaneous Localization and Mapping) pointed to the possibility to determine one's own position and simultaneously to construct a map of the interior of the building. In the year, the system has been helm-mounted, developed towards realtime capacity, and been given considerably improved positioning capability through improved time synchronization between the IMU and the stereo camera. With a test system of our own construction we have continued to investigate the potential of our RF-based (Radio Frequency) "silent" positioning method of soldiers in buildings. In experiments performed at MSS Kvarn we have been able to verify the functionality of the system concept in buildings and its robustness against multipath propagation.