Project MagEye

Objectives of the project:

  • The main objective of the project is to develop a person detector, which will be able to detect, at a defined reliability rate, the presence of a person in the given scene regardless of the fact, whether he moves or not and in which position he is (standing, sitting, lying, ...)
  • The secondary objective of the project is such functionality of the presence detector which is able to count the number of persons appearing in the given scene

Strengths of presence detectors based on camera scanning:

  • The detector eliminates unpleasant feature of all commonly used presence sensors: for detection it is necessary for the object to move, at least to the minimum extent; if he is not moving, the sensor reports that there is no person in the observed scene
  • The detector reacts really only to persons (or objects corresponding by their shape to the shape of head and shoulders of a human); it doesn’t therefore raise alarm in cases when animals or other moving or static objects appear in the scene that are of different shape
  • By modification of identification programmes, it is possible to adjust the sensor for specific tasks (e.g. scanning of long closed space – corridors, scanning of entrance scenes, sensing with evaluation of the number of identified persons etc.)

Application field:

  • Energy saving (switching of lighting, heating and air-conditioning only in the presence of persons etc.)
  • Safety and supervision applications (monitoring of rooms where constant presence of at least one or more persons is required, or possibly where in a certain period of time no person is allowed to be present, while other static or moving objects may appear in the scene)
  • Special customized applications

Brief characterization:

  • Detection is based on models, not on detection of movement (the latter is used in auxiliary algorithms only)
  • Adaptive sensor: in the sensor’s software there are “learning“ algorithms applied, which, upon installation of such sensor in a new environment, at first „view the scene“ and afterward they use such information in the proper detection of persons
  • The sensor detects head and shoulders as distinctive features of a person, while the remaining parts of human body may be hidden and are not of substance for proper decision making
  • Recognition algorithm is not based on recognition of person’s face
  • The sensor doesn’t register in any way persons entering the scene and leaving the scene; the detection algorithm may start at any condition of the scene, however, it doesn’t register hidden persons (not visible for camera)

Hardware configuration:

  • Miniature PC with operating system on the basis of Linux
  • Version for mounting on the wall or under the ceiling according to the grade of coverage, for indoor and outdoor spaces
  • Taken scene is defined by the camera optics used (focal length) and by its setting-up
  • The sensor contains a relay output which may be used for switching energy or lighting sources (lighting, heating), various types of signalling
  • For the purposes of setting-up and for transfer of information on the number of identified persons, there is a Wi-Fi output
  • The sensor is equipped with IR RED diodes for sensing of scenes in dark or under poor lighting