E-mail: inquiry@miaxis.com

Biometric Products

Home > Biometric Products > Biometric AFIS > Face Recognition

  • Face Recognition Algorithm

Product Name: Face Recognition Algorithm

Share
Overview:

Miaxis's Justouch® facial recognition technology is one of biometric recognition technologies based on processing information of facial features, mainly used to automatically detect and track face then process each faces image in still images and video streams that captured by webcams or cameras and identify user's identity.


Available as a software development kit that allows easily integration into final products or biometric systems to realize identity management, national ID card verification and other functions based on advanced face recognition technology. Widely applied to smart finance,intelligent security scenarios.


Features:

●Face detection&tracking

Quickly and accurately detect the face and track location of each face in still image or video streaming


●Accurate Localization of Facial Feature Points

Accurately locate the key feature points on the face in the picture (5 key points such as eyes, nose, mouth, etc.), used to determine the face posture and face alignment


●Face image quality assessment

A quality threshold can be used during face enrollment to ensure that only the qualified face template will be stored into database to decrease the effect of image quality on matching performance.

 

●Live face detection

A conventional face identification system can be tricked by placing a photo in front of the camera.

We are able to prevent frauds with using photograph or 3D printed mask with SDK as follows,

1) active mode with face images captured in visible light: SDK requires user actively assist to perform particular actions such as open mouth, head movements to judge it's a 'live' human or not.

2) Near-infrared version: using imaging features of Near-infrared to achieve high robustness of live face detection.

 

●Face identification capability

It can be used for fast matching in 1:1 mode (verification), as well as 1:N(n= large database size) mode (identification).  Verify if it's same person by calculates two faces's similarity based on features extraction. If the matching score is higher than the pre-defined facial matching threshold then the input template is said to have successfully matched with the stored template. 

Face Recognition Algorithm

Specification:

Justouch® facial recognition algorithm is capable of fast processing speed,high accuracy and easy to integrate, available as a standard and customized version software development kits for windows, android,linux platforms to support our industry partners and system integrators.

A. Enrollment requirements
The facial authentication algorithm is based on the face recognition from two-dimensional images or videos. The following corresponding basic requirements for images and videos
Angle toleranceHead Roll<35°  Head Tilt<20°  Head Yaw<30°
IlluminationUniform illumination, can be identified if backlight and exposure are not seriously bad
BlurThe contour of the face is clearand the locations of facial features can be recognized.
Integrality It is necessary to ensure the integrality of the face, and the contour of face is all in the picture;and make sure the key parts of the face such as the eyes, nose and mouth would not be obstructed by any subjects.


B. Face algorithm performance
Template extraction speed(a single template) with a 640x480 image sizeWindows withi5-8250U processor≤100ms
Android with RK3288≤ 300ms
Min. Image size≥ 35 pixels between eyes for best performance
Matching speed(1:1 mode)Windows withi5-8250U processor≤0.01ms
Android with RK3288≤0.1ms  
Template sizeas small as 4 Kilobytes
Algorithm AccuracyFAR ≤0.01%

FRR ≤1%(Still image 1-to-1 matching)

FRR ≤2%(Still image 1-to-1 matching against image stored onto person's ID card)

FRR ≤5%(1-to-N matching against the whole database,N=5000)

Multi platformsWindows,Android,Linux



Case Study:


Face Recognition Algorithm


Face Recognition Algorithm