The difference between face capture and face recognition

Core differences

Comparison mode

• 1:1 Verification: Confirm "are you who you claim to be", such as facial recognition access control and payment verification. It has fast speed and high accuracy, and the threshold can be set higher.
• 1:N identification: Find out "who you are" among N people, such as blacklist control. The larger N is, the higher the similarity threshold and computing power requirements are, and the greater the risk of false positives.

Factors affecting accuracy

• Face pixels: The recommended distance between the eyes is ≥ 60 pixels. Below 40 pixels, the recognition rate drops sharply.
• Posture: The front face is the best, but the accuracy drops significantly for side faces at large angles (>45°).
• Lighting: Backlight, strong shadows, and no fill light at night will all affect it.
• Occlusion: Masks, hats, and glasses will cause feature loss.
• Quality of the library: Photos of the library must be clear, front-facing, and recent.

Hardware requirements

• Specialized face capture camera (large aperture, wide dynamic range, low illumination), which is different from ordinary surveillance cameras.
• The recommended installation height is 2.2 to 2.8 meters, and the depression angle is ≤ 20° to ensure that the front face can be photographed.
• The width of the passage should be controlled between 1.5 and 2.5 meters to prevent multiple people from entering the picture at the same time.
• The fill light should be soft and even to avoid yin and yang faces.

Privacy and Compliance

• Prior to collection, it should be publicly informed and clear signs should be set up.
• Follow the minimum necessary principle and do not collect in non-essential areas.
• Data is encrypted and stored, access rights are strictly controlled and retrieval logs are retained.
• Clear the retention period and automatically delete it upon expiration.
• When it comes to face collection in public places, you need to pay attention to local regulations and requirements.