The new VideoEdge facial biometrics feature immediately detects and verifies individuals. VideoEdge facial biometrics engine has been tuned to be fast and accurate. Whether you are using it for identifying known unwanted persons in real-time or providing secondary verification with access control, time is paramount. As with any analytics feature in video surveillance software, there are best practices that should be used to achieve reliable results and improve performance.
Recommended Camera Location
To recognize a face, the person should be easily visible. Consistent, accurate facial recognition results require the cameras to be mounted within a close range.
- The entire face should be visible inside the image and region of interest.
- Both eyes should be open and visible.
- Limit the number of faces in the scene. Try to avoid crowds and large groups of people.
- Reduce subject and camera motion as much as possible.
Provide Ideal Lighting
American Dynamics facial biometrics technology matches multiple points on a face. Therefore, the clearer the face image is, the more likely you are to receive an accurate match. You can resolve many of the lighting conditions necessary to achieve optimal facial biometrics by following these considerations:
- Create even lighting on the face with little to no shadows.
- Avoid backlighting that may result in glares on glasses. Position light sources outside the area where reflections hit the camera.
- Keep away from direct sunlight when possible. Direct sunlight may cause strong, asymmetric lighting and shadows across the face.
- Utilize automatic gain control in environments with varying brightness.
Upload Quality Enrollment Images
American Dynamics recommends you upload at least five enrollment images of each individual to the database. These images include varying angles such as straight on, face slightly turned left, face slightly turned right, face slightly tilted up and face slightly tilted down. Having each angle should improve the recognition performance. Also, you should use recent enrollment images and remember to re-enroll individuals periodically. Often times, age differences or changes in hair of facial style has a direct effect on the recognition performance.
- Use a high resolution image with at least 400 pixels on the face
- Take enrollment images with neutral facial expressions including mouth closed and little to no smile.
- Take the image close up, recommend 3 to 5 feet away
View our best practices cheatsheet.
Learn more about American Dynamics VideoEdge Facial Biometrics.