How it works
Sigur uses facial recognition to automatically identify cardholders at access points. The facial recognition feature captures the cardholder’s face from the video and compares this picture with the database of cardholders’ photos stored in the ACS.
The video stream can be obtained from IP cameras directly connected to the system. The same cameras can be used as a video surveillance system of the site.
Supported modes
Identification In this case, the cardholder’s face is the only parameter that the system relies on in whether to grant access or not. It is the most convenient mode as it does not require any additional actions on the part of the cardholder.
Verification In this case, this biometric parameter serves as an additional identification means. Other credentials such as a proximity card can serve as the primary identification means. There are two types of the verification mode:
- «Strict»: after the identification using the primary identifier, the system will compare the image obtained from the cameras with the cardholder’s photo. If the face has not been recognized or the cardholder has failed to appear on the screen within 5 seconds, the system will deny the access.
- «Hard group verification» is a new verification feature in Sigur. When activated, this feature will, after the recognition of the primary credential, proceed to matching the face of the employee with one of the faces from the same group in the system to which this particular employee belongs.
- «Soft»: the system will provide access in any case; however, if the face has not been recognized, this event will be prompted in the monitoring interface.
The system provides flexible configuration options in the identification mode with the option to set up:
- individual access points and directions
- specific groups of cardholders
Operating requirements
As with any other biometric identification technology, facial recognition is not 100% accurate and has a number of limitations of use. The identification mode will operate accurately only with a limited number of faces in the PACS database—up to 1,000. The accuracy of the system also depends on:
- face positioning relative to the camera
- quality of the photos and video stream
- quality of lighting at the scanning area
- face image resolution (min 100px)
System requirements
For the recognition system to work, any up-to-date PC and any IP camera will suffice. E.g. the system operating in a 4-channel mode requires Intel Core i7 and at least 8GB RAM.