The importance of optimal camera placement for face recognition and video analytics
As many of the first-generation IP camera systems reach the end of their useful lives and are due for replacement, it's critical to ensure that new deployments are designed for optimal analytics performance. This is especially important given the rise in the use of analytics in general and the growth of surveillance systems in general.
Camera design ideas such as using pixel-per-foot calculations to guarantee acceptable details are attained, or using a lux meter to pick the camera with the appropriate low light capabilities for a scene are generally common to systems integrators and bigger end customers.
Designing for optimal analytics performance, on the other hand, is rarely considered, especially if analytics aren't part of the initial system deployment. This guide is intended to offer advice for ensuring that your security camera designs are suitable for analytics use cases now and in the future.
Ultra Wide Angle Lenses
In recent years, cameras with 180 or 360-degree lenses have grown increasingly popular, particularly for interior applications such as corporate lobbies and common areas of commerce. However, because the image geometry is not uniform, these cameras frequently produce images with significant distortion, which can often hinder analytics performance.
Distance to Subject
Facial recognition at extremely long distances can be tricky for various reasons. It is possible but comes with the cost of buying a much more sophisticated and expensive camera model.
Using the most common wide-angle lens with a focal length of 3.6 mm, the optimal subject distance for accurate identity identification will be anywhere from 5 to 20 feet (1.5 to 5m).
However, with today’s wide range of lens options, it is possible to capture sharp subject details even over long distances. Disadvantages of extremely long-distance shoots include small vibrations or movements at the camera becoming amplified over distance, making the image unstable. And the increased possibility of the subject coming between the camera and the far end of the field of view, thus blocking the camera's view. Furthermore, weather conditions such as fog or rain can reduce the operating range of the camera in outdoor applications.
As an example, if you are using a typical 2MP camera with a 1920x1080 pixel resolution and want to detect objects such as a human body or a car, at a distance of 50 feet (45m), this would require a lens focal length of around 7.41mm, depending on the camera model specifications.
And If you want to accurately identify faces at this distance, it would require a focal length of the lens of at least 18mm.
For identity identification, limiting the camera-to-subject distance to 50 feet (15m) or less will provide the most flexibility in camera sensor and lens combination options and reduce the chances of interference from camera vibrations or other environmental conditions affecting scene quality.
Understanding camera positioning requirements is crucial. Make sure the design takes these constraints into account, and that you can afford to make any necessary changes. As a result, the chances of an optimally functioning system with minimal surprises will be maximized.
Optimal camera tilt angle
A tilt angle of 15° to 45° will provide the most reliable results for general object detection, such as body or head detection.
However, for identity identification, a tilt angle of 0° to 15° will be more effective, giving the camera the capacity to capture a clear view of faces.
Optimal camera height
The best installation height for facial recognition is 6 to 9 feet (1.8 to 2.6m), allowing for minimal downward tilt angles. More direct views of the subject's faces are possible, which improves identification capability.