Facial recognition module
The facial recognition module is used to automatically identify people by their video images. It recognizes faces captured by Axxon facial detection tool by comparing their parameters with digital templates stored in a dedicated database. In addition to automatic identification of people based on video images, the module enables you to: add or remove entries from the reference database; print photos of recognized faces or export image files in .bmp or .jpeg format; search the database and display recognition statistics; review video recordings related to specific face recognition events; check images’ compliance with standards for biometric identification systems. The facial recognition module ensures high accuracy of recognition and can be used in combination with an access control system on sites with higher security requirements, such as banks or restricted access facilities. Another important application for th module is automated face control in casinos, hotels, restaurants, or other similar facilities.Principle of operation
The facial recognition module works with a camera and the Intellect facial detection tool. Firstly, the facial detection tool detects the presence of a face within a video frame and captures its image. The face recognition module has two possible operating modes: identification and verification. In identification mode, the parameters of a captured face are compared to all facial templates in the database. This helps determine, for example, whether or not the person is blacklisted or a welcomed VIP customer for the particular facility. In verification mode, the face of an access card holder or a person using another sort of ID to gain access through a turnstile or restricted access door is compared with the rightful pass owner’s photo stored in the database. This helps verify that the person requesting access is actually the person he/she claims to be. The module’s settings let you define the levels of similarity (expressed in percentage) that form the boundary limits of what are known as similarity zones. The following three zones can be configured: red (high similarity), yellow (medium similarity), and green (low similarity). If similarity is high, the recognized face is saved to the facial database along with the date and time of recognition, camera ID, and the degree of similarity. To simplify monitoring, the similarity level is color coded on operator’s displays. In addition to recognition, the module lets you delete existing records from the template database. You can also create new records containing images, their parameters and personal data, such as full name, department, relevant comments, etc. A reference image can be any digital photo imported into the database, or any image captured by cameras in operation. The module also lets you verify images’ compliance with international standards for automatic personal identification in biometric systems (ISO 197945).Facial recognition module interface
The module’s interface consists of three components, each performing different tasks: a screen displaying live feed from the camera, a facial monitor, and a recognition monitor. The facial monitor displays the most recent images of faces captured by a camera. The number of images is determined by window size settings. Each image is date / time stamped with the time of capture and camera ID. In case it matches a database,item, the degree of similarity and the person’s name in the DB are also given. A separate window on the face monitor displays the most recent recognized face, the matching image from the database, full name of the person on records, date/time of recognition, relevant comments, and similarity percentage. You can also view the source video sequence where the face was captured. The recognition monitor operates in two modes: Report and Archive. In Report mode, the screen displays data pertaining to the recognized faces, including captured image, database reference image, camera ID, similarity rate, full name, date and time of recognition. In Archive mode, the recognition monitor lets you retrieve a record from the database by following parameters: time of recognition (a specific time span can be set), camera ID, person’s name, and similarity rate (within a specified range). The retrieved faces are displayed in the same manner as in Report mode.Features
Faces are recognized in images captured by the Intellect facial detection tool by comparing image parameters with database templates. In identification mode, images are compared against all items in the database. In verification mode, the face of a person requesting access (using a proximity card, or biometric identification, etc.) is compared to the corresponding image in the database. Comparisons are based on Cognitec‘s face recognition technology. Captured faces are displayed on-screen along with relevant data, including date/time of capture and camera ID.
Recognized faces are displayed in the same way as captured with an addition of database reference photo, full name of the person, similarity rate, and arbitrary comments entered by operator. of recognized persons can be printed out along with personal data, or exported in .bmp or .jpeg files.
Video sequences corresponding to specific face recognition events can be reviewed.
All faces recognized within a specified time frame can be searched and displayed on screen. Searches can be performed by full name and camera ID while filtering for similarity rate.
Searches can be performed also by database reference photos.
New database records can be created containing digital images, personal data and arbitrary comments.
Images in the database can be checked for compliance with international standards for biometric identification systems (ISO 19794 5). Compliance checks can be performed both on single images and on the entire database. Reports are saved and subsequently displayed whenever the particular image is viewed.
The following recognition statistics are displayed on screen: total number of frames captured by the facial detection tool, number of frames on which the face was detected by the recognition algorithm, number of frames on which eyes were detected, and total number of recognized faces within an image.