How to figure out the best distance, height and angle for the job.
First, we need to ask ourselves a few questions. What are we trying to accomplish? What problem are we trying to solve. Are we looking for higher detail for better identification and recognition? Are we more interested in analytics like detecting objects such as people or vehicles? And, does it need to be an either / or scenario?
Let’s dive in…
What is pixel density and why does it matter? In short it is the number of pixels used for a given object within the field of view. Pixel density combines the sensor size, pixel count, focal length and distance to the object into just one parameter. The FoV is commonly described as a measurement such as Feet or Degrees. Pixel density is usually expressed in PPM (Pixels Per Meter) or PPF (Pixels Per Foot).What about resolution?
Let’s look at some pretty common resolutions for security cameras. Although 4k is still not that common because the cost, bandwidth, and storage for the video stream can be cost prohibitive. We will pay extra attention to the horizontal pixels. We will use these in our PPM and PPF calculations later.
| Resolution | Horizontal | Vertical |
| 720P | 1280 pixels | 720 pixels |
| 1080P | 1920 pixels | 1080 pixels |
| 4MP | 2560 pixels | 1440 pixels |
| 4K | 3840 pixels | 2160 pixels |
Let’s look at the the FoV and FL of Meraki MV Cameras.
| Camera | FoV (Field of View) | Focal Length |
| MV12N | 73° | 3.8mm |
| MV12W | 114° | 2.8mm |
| MV22/72 | 36° – 112° | 3-9mm |
What level of quality do we need?
A couple of things to consider:
- The greater the pixel density, the greater the quality of the picture.
- The greater the pixel density, smaller details will be easier to distinguish.
The Department of Homeland Security published a Video Quality Handbook in 2013 and updated in 2018 which provides guidance on image quality requirements. In that handbook the following minimum resolutions were called out: (Note: DHS only uses PPF so I have added approximate conversions to PPM)
Observation – 66 PPM or 20 PPF
General viewing of scene content
Forensic View – 131 PPM or 40 PPF
General identification of scene content or object of interest; identification of object colors; and specific identification of an object’s characteristics, the time, and the location of the objects.
Recognition – 262 PPM or 80 PPF
Including but not limited to vehicle license plate recognition, facial recognition, face location, smoke and fire detection, object recognition, pattern recognition, cross-line detection, and color detection.
Please note that DHS is not setting some sort of requirement and they are not the only one publishing handbooks or standards. For example the European Standard EN 62676-4 dated from 2015 categorizes the following:
Monitoring – 12 PPM or 4 PPF
Think of this as being able to get overall situational awareness.
Detection – 25 PPM or 8 PPF
The ability to detect the presence of a human.
Observation – 62 PPM or 19 PPF
Ability to identify characteristic details of an individual, such as distinctive clothing.
Recognition – 125 PPM or 38 PPF
Someone viewing the footage should be able to recognize a person known to them. Viewers can verify with a high degree of certainty whether or not an individual shown is the same as someone they have seen before. Should be possible to recognize license plates manually, but probably not detailed enough to use an Automatic number plate recognition (ANPR) systems.
Identification – 250 PPM or 76 PPF
Positive identification of a person beyond reasonable doubt. With sufficient picture quality and detail to identify an individual.
Inspection or Strong Identification – 1000 PPM or 303 PPF
Now that we have some basic established specifications that we should try to meet or exceed we can match those up with our use case(s). We will then plug those numbers into some formulas to figure out the install height, distance, and angle from our target area.
And some final thoughts on what we have seen from both the DHS Playbook and the European Standard. Most commonly I am asked what is needed to identify a person and what is needed to identify a license plate. If we round up a bit it would be safe to say we should try to achieve a minimum of 33 PPM or 10 PPF when trying to detect the presence of a person. Second, if we want to be able to automatically identify license plates, facial features, and patterns we should be at 262 PPM and 80 PPF.
Now what about the angle? What is the correct or best angle for the camera to capture what is important? That depends… If object of interest or target is a person and more specifically that persons face, then you will want to have the camera angle at less than 30 degrees.
Now let’s put this all together!
I think it best to start this out with the formulas and some examples.
Pixel Density = Horizontal Pixels / Focal Width at the object (FoV Width)
Simple right…? Well, not so much. It really depends on the distance between the object of interest and the camera. The closer the object is to the camera the better the pixel density. Remember the object of interest exists in 3-dimensional space but will be displayed on a 2-dimensional image.
FoV Width = Distance of object * Sensor size in mm / Focal Length
1/4 sensor = 3.2mm
1/3 sensor = 4.8mm
Let’s look at a simple example using a Meraki MV12n Camera.
Note: Meraki MV Cameras use a 1/3″ image sensor
Example 1
Object of interest distance from MV12n = 15ft
Focal Length = 3.8mm
Horizontal Pixels @ 1080P = 1920 pixels
FoV @ 15ft = 18.9ft
Pixel Density = 1920 / 18.9ft
~333 PPM or ~102 PPF
Example 2
Object of interest distance from MV12w = 15ft
Focal Length = 2.8mm
Horizontal Pixels @ 1080P = 1920 pixels
FoV @ 15ft = 25.7ft
Pixel Density = 1920 / 18.9ft
~245 PPM or ~75 PPF
Before we continue with more examples I want to drive home one key piece of the formula from the first 2 examples. Specifically the distance from the camera to the target of interest.
Let’s consider the following scenario posed to me by a customer. They wanted use a security camera to do people counting, gender identification. They asked is this possible, and if so where do we install the camera? My answer was yes this is possible but if we get the install location correct. I asked about any other requirements. They said they wanted the camera installed on the ceiling. We have a camera installed on the ceiling at a height of 12ft. We want to monitor the people that come through the entry door(s) to a retail location. We need to figure out what quality we want to achieve. Once we have that figured out we can start playing around with the numbers. When people walk through a door we may be interested in several things. It could be that we just want to view a basic level of detail of the scene, it could be we want to identify if a person walks through the door, and it could be that we want to positively identify who that person is. Those are 3 very different set of requirements.
How do we calculate the distance between a ceiling mounted camera and someone that just walked through an entry door? Let’s answer that with a specific example.
Example 3
MV12w mounted on the ceiling at a height of 12ft
The camera is installed 20ft from the entry door
The point of capture is 3ft inside the entry door
The average height of the centerline of a persons head as they enter the capture zone is 5ft
Using this information we can use math… to calculate the distance of a persons head in the target area.
**Please reference the diagram below to help visualize this example**
(a) Height of Camera = ceiling height – target height
7ft = 12ft – 5ft
(c) Capture Zone = entry door distance from camera – 3ft
17ft = 20ft – 3ft
(B) Angle in degrees from floor to Camera at install locations = 90°
(A)22.38° is the angle from a persons head to the camera
Based on the distance of 18.385ft from the target the the FoV Width is 31.5ft. From this we calculate a 200 PPM or 61 PPF at the target of interest.
What about height and angle?
In order to capture the greatest amount of details from a person / targets face we need to keep the angle (A) noted in the above diagram between 15º to0 30º. Ideally as close to 15º as possible. Please keep in mind that most people / targets will have their heads pointed straight ahead at 0º or at a slight down angle up to 20º. Therefore it is very important to have the camera installed at a height that allows for the adequate capture of details for persons / targets with varying angles of approach.
Links to websites used for calculations and reference.
https://documentation.meraki.com/MV/Video_Analytics/MV_Object_Detection
http://www.golbong.com/Quality-image-identify-face-with-CCTV-camera.asp
http://www.jvsg.com/pixel-density/
https://documentation.meraki.com/MV/Viewing_Video/Understanding_Image_Quality_on_the_MV32
https://www.ultimatesecurity.com/fov-calculator
http://www.endmemo.com/sconvert/pixel_mpixel_ft.php
https://documentation.meraki.com/MV/MV_Overview_and_Specifications/Second_Generation_MV_Cameras%3A_Overview_and_Specifications
https://www.aqua-calc.com/calculate/image-resolution
https://www.calculator.net/triangle-calculator.html
https://www.dhs.gov/sites/default/files/publications/VQiPS_Digital-Video-Quality-HB_UPDATED-180117-508.pdf