Note: FaceR functionality is not available in VDG Sense 2.5 or higher
Face recognition can isolate faces from video images and convert them into metadata. Metadata points are placed on eyebrows, eyes, nose and mouth. Each recognized face is compared with the pattern stored in the database and on the basis of the so-called automatically classified. black or white list closely associated with root access to the resources of the object whose face is analyzed. Based on the results of that analysis, the system activates the appropriate rule macros. Activating a dedicated profile allows you to verify the presence of people in the space object specifying the time period.
Face Recognition should only be used indoors in quiet areas where people are required to look into the camera. A practical application could be access control of server rooms whereas a limited number of people have access to that area.
Requirements of camera mounting spot
Use of artificial lighting to provide adequate illumination face (preventing shadows)
Avoid light that creates shadows
Camera position should be that people are encouraged to look straight into the camera
The face should be completely visible, nothing should be covered.
The distance between the eyes must be for a minimum of 60 pixels for optimal performance, higher is better.
Max camera horizontal and vertical viewing angle of 15 degrees
FaceR image examples
Score: 1.00
Properly illuminated
Score: 0.94
Glare from lighting, Test with people wearing glasses
Score: 0.65
Light source from above;
causes shadows.
Score: 0.32
Light source from above;
different pose
Score: 0.41
Properly illuminated,
but also lateral daylight.
Score: 0.03
Strong sunlight, over-exposure.
Worst scenario.
Limitations of the algorithm
Can only be used indoors
Can only be used in quiet areas
Sunglasses affect performance
Shadows in the face affect performance, for instance light from above or from the side
Low contrast/low lighting affect reliability
People are required to look straight into the camera