Face Detection vs Face Recognition
Face detection is used to detect if a human face appears in the video stream. The camera or recorder's advanced algorithm can then capture the face, record, activate snapshot, and activate an event or alarm triggers. The detected face's attributes are also extracted and can be searched via metadata.
Face Recognition will not only detect if a human face appears in the video, but also identify the person's face, and compare that face to a created database. If a match is found, the system will notify you of the person's identity and trigger any events or alarm triggers. For some supported systems, a detected face that does not resemble any faces in the database can be tagged as "Stranger," and an event or alarm trigger can be activated.
When scoping or bidding on a project, it is important to keep these differences in mind. The cost difference between the two can be significant depending on the client or project requirements.
- Face Detection and Recognition requires a face to be at least 150x150 PPF (Pixels Per Foot) image in the video. This means that for Face Detection to work properly, a person's face must take up at least 150x150 pixels, and the distance between the center of both eyes must be more than 50 pixels.
- The larger the face image (in pixels), the more accurate it can be compared to the faces in the database.
- The camera or recorder can theoretically capture an unlimited amount of faces if they meet the requirements above.
Example of Face Detection:
Example of Face Recognition:
Example of Face Recognition Stranger Mode (if a face detected is not in the database):
Face Detection Requirements
The horizontal distance between the camera and face detect target is d(f), the camera installation height is h(f), downward angle of the camera is α(degree) (it is the included angle between the camera monitoring direction and horizontal surface). You can refer to the following parameter table:
Parameter | Range | Recommended value |
---|---|---|
Height h(feet) | 6.5~9.8 | 8.2 |
Horizontal distance d(feet) | 13.1~65.6 | 19.6 |
α(degrees) | 32.8~49.2 | ≤15 |
Face Detection Lens Recommendations
Lens |
Description |
2.8mm | Not generally recommended |
3.6mm | Not generally recommended |
4mm | Ok |
6mm | Good |
8mm | Better |
Varifocal | Best |
Fixed Lenses
A camera's Face Detection performance is largely dependent on having a larger lens. A camera with a longer focal length is better (a narrower view, but much better performance), while a varifocal lens is best.
In this regard, a 2MP fixed camera with an 8mm lens will be more effective in detecting faces than a 4K fixed camera with a 4mm lens. A 6mm or 8mm lens is strongly recommended if varifocal cameras are not an option.
Varifocal
When scoping a project, ask the customer what their requirements are for Face Detection. If they require both Face Detection and a wide angle of view, consider using two cameras for the job. One varifocal for Face Detection, and one fixed lens for overwatch of the area.
Varifocal cameras will give you more flexibility during testing, and for making minute adjustments after the camera has already been installed. If a fixed camera is having consistent issues capturing or detecting faces, a varifocal camera will be the best solution, but at a higher (up to 3 times) cost.
Comments
0 comments
Please sign in to leave a comment.