Testing Camera Height vs Image QualityBy: Derek Ward, Published on Sep 27, 2013
A critical debate amongst surveillance professionals is how high should cameras be mounted. The higher the camera is mounted, the less likely it is to be vandalized, stolen or obstructed. But the higher it is mounted, the more likely image quality suffers as the angle of incident steepens (as shown to the right).
The recommendations vary tremendously. Many say never to mount cameras less than 12 to 14' off the ground (4+ meters). Others say, the camera needs to be as close to human height as possible to get full facial details.
In this report, we put these two competing positions to the test with a series of experiments.
We mounted a camera to an extended pole on our test rig, allowing us to adjust camera height from 17' to 5', as shown below:
We then positioned a subject at short, mid, and long ranges, in order to see the difference mounting height and resulting angle of incidence made in details captured.
Moreover, we did tests with a subject looking straight ahead as well as walking with their head tilted down, since many people naturally do so, or worse, try to avoid their face being captured on camera.
The report answers the following questions:
- For a subject 6', 24' and 48' from a camera, what is the maximum height a camera can be mounted while delivering clear facial details?
- How much of a practical difference does a person looking downward have in selecting optimal camera height?
- What is the maximum downtilt of a camera that allows a complete face to be captured?
- What does a subject look like from different camera heights at the same position?
Here are our key findings from this report:
- ~15° is the maximum angle of incidence which should be used to capture facial details
- At close ranges, camera mounting height is crucial. Even mounted at 7-8', cameras begin to provide decreased facial detail of close subjects looking straight, due to the increased angle of incidence. Above this, the issue is worsened, with the camera capturing the tops of heads.
- Considering a natural 15-20° head tilt of a walking object, cameras should be mounted even lower to less than 6' high for a full face capture.
- Even a 15-20° tilt of object head required the camera to be mounted 2'-3' lower to get the same facial details as a setup that assumes subjects will walk with their head straight.
- At medium ranges, mounting height's effects are quickly lessened. At 24' from the camera, there was only modest difference in image quality due to angle of incidence, with no practical negative impact.
- At long ranges, increased mounting height has little effect, as the difference in camera angle is minimal.
This image comparison shows the differences in a real life door entrance scenario:
Though industry common knowledge says cameras should be mounted "out of reach" of vandals, if targets will most often be at close ranges, this practice will be of huge detriment to details delivered, meaning the difference between capturing identifying facial details or not. Cameras should be mounted as low as is practical to avoid this.
For longer distances to target, mounting height has little impact, and cameras may be mounted higher without practical ill effects.
Recommended Maximum Mounting Height
Based on our close range test, ~15° is the maximum angle of incidence which should be used to capture facial details. Above this, facial details are lost, with only the top of the head being captured.
Using this critical angle, the following table lists the maximum mounting heights recommended to capture facial details of a 6' tall person at varying distances. Note that camera height listed in the table is a maximum value. Cameras may be mounted lower as well. Also note that the consideration of a small tilt of object head significantly reduces the maximum camera height. We recommend the use of a camera height that is about the same height or less than the height of the object to capture full facial details of an object with head tilted down.
If calculating for a shorter object, for instance a car number plate (~2' off the ground), users should subtract the difference from the maximum camera height in the table above. The same is true of taller objects; simply add the difference.
Note that this table should be used as a starting point only, as it does not consider other factors which affect image quality, such as lighting, resolution, WDR performance, etc.
Angle of Incidence
Before moving on to test results, angle of incidence should be briefly mentioned. It is an important metric that affects the ability to recognize object details and is purely dependent on the object distance, object and camera height. An illustration of angle of incidence is given below. When the angle of incidence is small, the details are more detectable. As the angle widens, they become more difficult to see.
Close Range (object at 6')
At close ranges, 6' away from the camera in our case, the camera height has the most critical importance. The facial details of the object, who is looking straight ahead, were clearly visible when the camera was mounted at a lower position, from the minimum test height of 5' to about 7'-8' high. When the camera was moved up to 10'-11' high, the facial details became difficult to distinguish due to increased angle of incidence. Further above this height, only the top of object's head was visible, with no facial recognition opportunity.
To better observe the differences due to varying angles of incidence, see this animation below.
A proper mounting height for facial recognition is also dependent on the face direction of the object. When people are walking, they sometimes tend to tilt their head downwards, generally with an angle of around 15-20°, instead of looking straight. Even this small tilt has a significant effect on the angle of incidence, especially at close ranges.
We tested this effect with a similar approach, but with object tilting his head slightly down. This time, angles of incidence will be much higher for the same mounting height.
Here are the results at various mounting heights and in comparison with the object looking straight ahead.
When the object face is slightly downwards, facial details started to diminish just after our minimum mounting height of 5'. These results show that even a 15-20° tilt of object head required the camera to be mounted 2'-3' lower. Only the top of object's head was visible for this case, when the camera was mounted at 8' or higher.
Based on a 6' object, 6' away, we can determine the angle of incidences for various height and object face directions, as follows:
When the object looks straight ahead:
- For a camera height of 6', the angle of incidence is 0° (direct incidence, best detail capture).
- For a camera height of ~7'-8', the angle of incidence is around 15°. (The widest angle where facial details are still visible, i.e. critical angle).
- For camera height of 10', the angle of incidence is ~35°. (Above critical angle, eyes are not seen at all but nose can be barely seen).
- For a camera height of 17', the angle of incidence is around 61°. (Very wide angle, only top of object's head is visible).
Since the object is very close to the camera, the angle of incidence is very sensitive to the camera height. The angle is varied between direct incidence to about 60°, with a height variation from 5' to 17'. The critical camera height, or the maximum mounting height that can capture facial details, was observed to be around 7'-8' for this case.
When the object head is tilted downwards:
- For a camera height of 5', the angle of incidence is around 10° (Below the critical angle, face details can be captured).
- For a camera height of 6', the angle of incidence is around 20° (Slightly above the critical angle, face details start to diminish).
- For a camera height of ~7'-8', the angle of incidence is around 35°. (Above critical angle, face partly visible).
- For camera height of 10' and higher, the angle of incidence is ~55° and higher. (Very wide angle, no facial details at all).
Slight downward tilt of the object head increases the angle of incidence. Above the minimum test height of 5', facial details started to be lost.
Mid-Range (object at 24')
At 24', mounting height has a much less pronounced effect. Only above a mounting height of around 11'-12', the angle of incidence passed the critical angle of 15°. However, even at the maximum height of 17', the angle was still not larger than 25°. The below picture compares the images from three different mounting heights.
The following animation shows these differences as the camera moves down from 5' to 17':
The angle of incidence ranged from 0° at 6' to about 25° at 17' camera height. Since the angles are much smaller than those in the close range, image quality does not change as significantly.
At 48', the effect of mounting height was minimal. All tested camera heights generated almost the same image, with facial details detected similarly. The maximum angle of incidence at 17' high was only 13°, still below the critical angle. Details delivered in this case are reduced due to much lower PPF than in prior ranges, not mounting angle.
This animation shows the difference in details as the camera is moved from 5' to 17':
The maximum angle of incidence obtained at 17' is around only 13°, which is an indication of how small the mounting height affects the image quality for distant objects.
To show a typical real world scenario, we set up the camera near an overhead door and observed our subject walking in, nearly perpendicular to the camera. The image shows the difference in details captured at several heights, with the subject's head tilted slightly down. As we found in other tests, facial details are significantly lost with camera heights above 6'-7'.
Users may download clips of this scenario to see differences for themselves.