Detect, Observe, Recognise or Identify? CCTV Field of View and Coverage Design under the SPF VSS Standard

A CCTV camera can cover an area without actually showing enough detail to achieve its security purpose.

This distinction is fundamental to good surveillance design.

A camera may clearly show:

“There is a person at the entrance.”

but still be unable to show:

“Who is that person?”

The Singapore Police Force Video Surveillance System Standard for Buildings addresses this by defining four different levels of surveillance detail:

Detect → Observe → Recognise → Identify

Each level represents a different operational outcome.

The Standard then applies different levels to different building locations. General common areas may require Observation, while frontal views at main entrances and exits are expected to achieve Identification.

This leads to one of the most important principles in CCTV design:

A camera should not simply have “coverage”. It should provide the required level of detail at the location where that detail is needed.

What are Detect, Observe, Recognise and Identify?

The SPF VSS Standard uses four categories to describe the amount of visual detail required from a surveillance image.

Detect

The purpose is to establish whether a person is present in the scene.

The Standard’s traditional definition describes the person as occupying at least 10% of available screen height, in a scene that is not excessively cluttered.

At this level, the objective is essentially:

Is somebody there?

It is not necessarily possible to determine who the person is.

What does Observe mean?

Observation provides more information about the person and what is happening around them.

The Standard’s traditional definition places the figure at approximately 25% to 30% of screen height.

At this scale, characteristic details such as distinctive clothing may be visible while the scene remains wide enough to understand the surrounding activity.

The question becomes:

What is the person doing, and what is happening around them?

What does Recognise mean?

Recognition is intended to allow the viewer to determine with a high degree of certainty whether a person shown is the same individual the viewer has seen before.

The traditional definition places the person at least 50% of screen height.

The practical question is:

“Is this the same person I saw earlier?”

That is different from establishing identity beyond reasonable doubt.

What does Identify mean?

Identification requires the greatest level of detail of the four categories.

The Standard’s traditional description places the figure at least 120% of screen height, with sufficient image detail to establish the person’s identity beyond reasonable doubt.

The purpose is therefore:

Who is this person?

This is why Identification is particularly important at locations where a building needs a useful facial image of everyone crossing a specific security boundary.

Why does the Standard later give different percentages for 1080p CCTV?

This is an important point.

The original percentage-of-screen-height model was developed around older analogue video systems.

The VSS Standard explains that digital and megapixel systems can provide substantially higher image resolution, meaning the subject does not necessarily need to occupy the same percentage of the screen to achieve equivalent detail.

For this reason, the Standard provides equivalent percentage screen heights for different video resolutions.

For 1080p, its guideline values are:

The figures come from the VSS Standard’s digital-resolution conversion table.

So does Identification require 45% of the image height at 1080p?

The Standard gives 45% as the equivalent guideline figure for 1080p.

But it expressly warns that these values should only be used as a guideline.

Other factors still affect whether the image actually provides usable detail, including:

lighting, camera angle and image compression.

So the correct interpretation is not:

“45% automatically equals guaranteed identification.”

It is:

“45% is the VSS Standard’s guideline for equivalent Identification detail at 1080p, but actual image quality still has to be verified.”

What does 45% image height mean in practical terms?

For a 1080p image, a target occupying 45% of the vertical image would occupy approximately 486 vertical pixels.

That is a mathematical conversion of the Standard’s 45% guideline, not an additional requirement stated by SPF.

Similarly:

• 20% of 1080 pixels is approximately 216 pixels;
• 10% is approximately 108 pixels.

These calculations help explain why field of view matters so much.

If a person becomes very small within a wide scene, the camera may contain many total pixels but relatively few pixels are actually being used to show that person.

Is this the same as “DORI”?

In security-industry terminology, Detect, Observe, Recognise and Identify are often grouped together as a DORI-style field-of-view concept.

However, the SPF document itself presents the four categories by their full names.

For this series, it is more useful to work directly with the Standard’s terminology:

Detect, Observe, Recognise and Identify.

Does CCTV resolution alone determine the surveillance level?

No.

The VSS Standard specifically says factors such as:

lighting, angle of view and compression

also influence image quality.

It also says the resolution used in the comparison should reflect the lowest resolution in the imaging chain, not simply the resolution of the display monitor.

That is an important technical point.

What does “lowest resolution in the chain” mean?

A surveillance image passes through several stages.

For example:

Camera → transmission → recording → playback → display

If one part of that chain reduces image resolution or quality, the fact that another component supports a higher resolution does not restore the lost information.

Consultant’s perspective

For example, installing:

an 8MP camera and a 4K monitor

does not provide an 8MP evidential image if the recording configuration stores the footage at a much lower resolution.

The entire chain has to support the required outcome.

Can compression affect whether CCTV achieves Identification?

Yes.

The Standard specifically says recorded picture quality needs to be examined because compression can result in the loss of picture detail.

This is why simply calculating camera resolution and field of view is not enough.

The recorded footage must be checked.

Does the Standard assume a particular human height?

For its conversion guidance, the VSS Standard states that the person imaged is assumed to have an average height of approximately 1.64 m to 1.76 m.

This matters because percentage-of-image-height calculations depend on how large the target appears relative to the image.

What level of CCTV coverage is recommended for common areas?

The Standard says general access areas should have sufficient cameras to provide comprehensive coverage.

Examples include:

• main entrance lobbies;
• streets within the development boundary;
• pavements;
• car parks;
• loading and unloading bays;
• taxi stands; and
• bus stops within the development boundary.

General common-area coverage should achieve at least Observation level.

Why is Observation appropriate for general areas?

Observation provides a balance between:

seeing enough surrounding activity

and

seeing useful characteristics of people within the scene.

A common-area camera normally needs to help the operator understand:

What is happening in this area?

rather than necessarily identify every individual everywhere within the image.

Does “comprehensive coverage” mean every square metre must provide Identification?

No.

The Standard differentiates between general common-area coverage and targeted higher-detail views.

General common areas are assigned Observation-level coverage, while certain specific locations such as main entrances are assigned Identification-level frontal views.

This distinction is extremely important.

A good system may use:

wide Observation coverage + targeted Identification views

rather than attempting to make every camera identify everyone everywhere.

What CCTV level is required at main entrances?

The Standard says frontal views of people entering or leaving the building through main entrances and exits should achieve at least Identification level.

This is one of the clearest surveillance requirements in the document.

The camera should also be positioned looking towards the doorway, rather than simply looking down on the person.

Why is a frontal view important?

Because identification depends not only on image resolution but also on what part of the person is visible.

A very high overhead camera may clearly show:

hair, head shape and shoulders

while providing poor facial detail.

Consultant’s perspective

Where facial identification is the operational objective, the camera should be placed so that the expected direction of travel naturally presents the person’s face to the camera.

The entrance itself becomes an identification zone.

Does the entire lobby need Identification coverage?

The Standard does not say this.

It distinguishes:

general lobby/common-area coverage: Observation

from

frontal views at main entrances/exits: Identification.

This is a good example of why one camera may not always be enough.

A wide lobby camera may provide excellent Observation coverage.

A more focused entrance camera may provide the required Identification view.

Whether one camera can achieve both objectives depends on the actual design.

What about entrances along passageways, walkways or MRT-linked routes?

The Standard says frontal views of people entering or exiting via entry/exit points along:

passageways, walkways or MRT stations

should achieve a minimum Observation level.

This is different from the Identification requirement stated for the main building entrances/exits.

What CCTV coverage is recommended at emergency exits?

The Standard calls for frontal views in both directions of every emergency exit.

The entrances to emergency escape routes should also be covered by cameras from the public areas.

General views of these emergency exits should achieve at least Observation level.

This recognises that emergency exits can function as both:

life-safety routes

and

potential alternative security routes.

Why cover both directions of an emergency exit?

Because the security event may involve movement:

out through the emergency exit

or potentially

towards or through it from another direction.

The Standard therefore does not treat the emergency exit simply as another internal door.

What about vehicle entrances and exits?

The Standard says cameras at vehicle entrances and exits, including car-park gantry points, should capture the number plates of vehicles entering and leaving:

car parks and loading/unloading bays.

That is a different operational objective from general car-park Observation coverage.

Can the general car-park camera also capture number plates?

Possibly, but it should not simply be assumed.

A general car-park camera may have:

a wide viewing angle and an Observation-level objective.

Reliable number-plate capture may require a more controlled:

target location, camera angle, exposure and field of view.

Consultant’s perspective

The better approach is to treat:

“Monitor the car park”

and

“Capture the number plate at the vehicle entrance”

as two separate operational requirements, even if one camera can eventually satisfy both.

What level is required in car parks?

The Standard’s common-area coverage requirements place car parks and parking areas within general Observation-level surveillance.

At vehicle entry and exit points, additional number-plate capture is required.

So again:

general coverage and targeted evidential capture are different tasks.

What CCTV level is recommended for lifts?

For lifts acting as alternative entry and exit points into the building, the Standard says:

• frontal views of the lift doors for people entering the premises; and
• general views of the associated lift lobby

should achieve at least Observation level.

This section concerns the lift and lobby as access routes into the building.

What about staircases?

For stairs that act as alternative building entry or exit routes, the Standard similarly calls for:

frontal views of people entering

and

general coverage of the associated staircase areas

at minimum Observation level.

What level is required at security or registration counters?

For locations where people undergo:

security checks or registration

before being allowed further into the building, the Standard says frontal views should achieve at least Recognition level.

Examples given include checkpoints and ticket-issuance counters.

This is a useful intermediate requirement.

The camera should provide more individual detail than a general Observation camera, without the Standard assigning the same Identification requirement used at main entrances.

Why might a counter require Recognition rather than general Observation?

Because the individual is already interacting directly with the organisation.

The surveillance objective becomes more specific:

Can we later recognise the person who presented themselves at this checkpoint or registration point?

The surrounding environment remains useful, but more individual detail is required.

What are “Sensitive Rooms” under the VSS Standard?

The Standard gives an example of sensitive rooms as rooms containing important or critical VSS recording equipment, such as NVRs.

For these rooms, it calls for coverage outside the access door.

The general external access area should achieve minimum Observation level.

What happens if the Sensitive Room door has an intrusion alarm?

This is one of the strongest examples of system integration in the Standard.

Where the door is fitted with an intrusion alarm:

activation should automatically display the relevant camera image on the designated VSS monitoring display.

The general view of the door should achieve Identification level.

So the intended workflow becomes:

Alarm activates → relevant camera appears → operator verifies what is happening.

Why is this better than simply recording the camera?

Because CCTV can support real-time response, not merely post-incident evidence.

If an alarm is generated but the operator then has to:

search through a long camera list, remember which camera covers the door and manually display it,

valuable response time is lost.

Linking the alarm to the correct camera helps CCTV perform its operational role.

What are Critical Areas under the Standard?

The Standard gives examples such as:

air-intake access areas and rooftops.

For external access to these Critical Areas:

the surrounding access area should achieve Observation, while general door views should achieve Identification.

It also calls for internal CCTV coverage where Critical Assets are housed.

What level is required inside a room containing Critical Assets?

The Standard says general views of Critical Assets inside Critical Areas should achieve at least Observation level.

It defines a Critical Asset as a system, equipment or process whose destruction or damage could have a debilitating impact on the premises, such as a single point of failure.

This is therefore not simply ordinary room coverage.

What is the overall VSS coverage hierarchy?

A useful summary from the Standard is:

This table summarises Clauses 4.2 and Annex A of the VSS Standard.

Does the VSS Standard use Detect-level coverage anywhere in its key building-area table?

Interestingly, most of the principal building coverage requirements in the Standard are assigned:

Observation, Recognition or Identification

rather than merely Detect.

Annex A summarises the recommended levels for the various defined areas.

Detect remains an important Field of View category, but the key building areas summarised in Annex A generally call for higher levels of usable detail.

Does “Identification at the main entrance” mean every point within the camera image must be Identification level?

No.

The requirement is associated with the frontal view of people entering or exiting the main entrance.

Consultant’s perspective

This means it is useful to define a particular:

Identification zone

within the camera view.

As a person crosses that zone, the image should meet the required Identification performance.

The wider scene may simultaneously provide only Observation-level detail.

What is an Identification zone?

This is a practical consultant term rather than terminology defined by the SPF Standard.

It describes the particular part of a camera scene where the design is intended to achieve the specified identification performance.

For example:

The 1.5-metre-wide pedestrian path immediately inside the entrance door.

A person outside that target area may not occupy enough of the image for Identification.

The purpose is to make the design requirement measurable instead of vaguely saying:

“Camera covers entrance.”

Can one camera have different surveillance levels at different distances?

Yes.

A person close to the camera occupies more of the image.

As the same person moves further away, they occupy less of the image.

So within one camera view, there may effectively be zones where the available detail corresponds approximately to:

Identify → Recognise → Observe → Detect

as distance increases.

This is a consequence of field-of-view geometry rather than a separate rule stated by the VSS Standard.

Why is this important when drawing CCTV coverage cones?

A conventional floor-plan camera cone tells you:

where the camera can see.

It does not necessarily tell you:

what level of detail exists throughout that cone.

Consultant’s perspective

A stronger CCTV design drawing can therefore show:

• the full field of view;
• the target surveillance area; and
• where Identification or other required performance must be achieved.

This makes the design easier to validate during commissioning.

Does a wider field of view reduce target detail?

For a camera of fixed resolution, widening the scene means the available image pixels are distributed across a larger physical area.

Therefore, the target generally occupies fewer pixels.

That is why an extremely wide-angle camera may be excellent for situational awareness but poor for targeted identification at longer distances.

This follows directly from the Standard’s emphasis on image proportion and digital resolution.

Is digital zoom the same as designing for Identification?

No.

Digital enlargement after an incident cannot recreate image information that was never captured.

If a face occupied too few pixels in the original recording, enlarging it simply makes those pixels larger.

Consultant’s perspective

This is why the design needs to put sufficient image detail on the target during capture.

“Zooming in later” is not a substitute for the required field-of-view design.

Can optical zoom help?

Optical zoom or a narrower lens can place more of the camera’s available resolution onto a smaller physical area.

But the trade-off is a narrower overall scene.

This illustrates the normal design compromise between:

situational coverage

and

target detail.

The VSS Standard handles this by assigning different surveillance objectives to different locations rather than demanding maximum detail everywhere.

Why might two cameras be better than one high-resolution camera?

Suppose a lobby needs:

general Observation coverage

and

Identification of everyone entering the main door.

One camera may be able to achieve both, depending on geometry.

But another legitimate design could be:

Camera 1: wide Observation view of lobby activity.

Camera 2: tighter frontal Identification view at the entrance.

Consultant’s perspective

This can sometimes produce a more dependable surveillance outcome than forcing one very wide camera to perform two conflicting functions.

The decision should be based on the Operational Requirements rather than an arbitrary rule about camera quantity.

Does crowding affect CCTV coverage?

Yes.

The Standard specifically says common-area camera positioning should take into account periods of maximum human density and their effect on achieving the Operational Requirement.

This is easy to overlook.

A beautifully positioned entrance camera may work when the lobby is empty but become ineffective when:

people queue in front of it, several people enter simultaneously or temporary event structures obstruct the view.

Should CCTV design be checked with people actually in the scene?

The Standard does not prescribe a single commissioning test method in this section.

However, because coverage requirements are defined around the size and view of human figures, actual commissioning needs to verify that the required views and image quality are achieved.

It separately requires the agreed camera views and both live and recorded image quality to be documented at commissioning and periodically reviewed.

Consultant’s perspective

A practical test should therefore involve realistic subjects standing or moving through the locations where:

Identification, Recognition or Observation

is expected.

Should the test use only the live camera image?

No.

The Standard warns that compression can reduce picture detail and says recorded image quality needs to be examined.

The system should therefore be validated using:

live view and recorded playback.

The recorded image is particularly important because that is what will normally be used during a later investigation.

Should camera coverage be validated during both day and night?

The Standard’s field-of-view section does not prescribe one explicit day/night test procedure.

However, the wider Standard requires cameras to provide the specified picture quality under the applicable environmental and lighting conditions.

Consultant’s perspective

If an Operational Requirement needs to be achieved 24 hours a day, it is not enough to demonstrate it only at:

2 pm in ideal lighting.

The relevant view should be assessed under the conditions in which it is expected to operate.

What happens if an entrance camera has the right percentage but a poor viewing angle?

It may still fail operationally.

The Standard expressly warns that angle of view influences image quality and separately says entrance cameras should look towards the doorway rather than simply downward.

So image-height percentage should never be treated as the only design criterion.

What happens if lighting is poor?

Again, achieving the nominal target size does not guarantee usable evidence.

The Standard specifically includes lighting among the factors affecting image performance.

A face occupying 45% of a 1080p image may still be poor for Identification if it is:

strongly backlit, underexposed, blurred or obscured.

What happens if compression is too aggressive?

Compression can remove image detail to reduce bandwidth and storage.

The Standard specifically says recorded image quality needs to be examined to make sure compression has not reduced important detail.

This is why designing by:

camera datasheet resolution alone

is insufficient.

How should a CCTV designer determine the required level for each camera?

Start with the operational question.

For example:

Do I need to know that someone is present?
Detect.

Do I need to understand what they are doing?
Observe.

Do I need to determine whether this is the person I saw earlier?
Recognise.

Do I need a sufficiently detailed image to establish who they are?
Identify.

Then determine:

where in the scene this outcome needs to be achieved.

Only then should the camera position and lens be selected.

Practical CCTV Field-of-View Design Method

For every camera, the design team should be able to complete something like:

This is a consultant’s application of the VSS Standard rather than a form contained in the Standard.

A practical main-entrance example

Assume the building entrance is 4 metres wide.

The client asks for one camera.

The first question should not be:

“Which 4MP camera should we use?”

Instead:

Requirement 1

Provide general awareness of activity around the lobby.

This corresponds broadly to an Observation objective.

Requirement 2

Provide a frontal image of each person entering the main entrance.

The Standard assigns an Identification objective to that view.

The designer then needs to determine whether:

one camera can actually satisfy both requirements

or

separate overview and targeted entrance views provide a better result.

A practical Sensitive Room example

Assume an NVR room is treated as a Sensitive Room.

The Standard calls for:

Observation of the external access area.

If the room door has an intrusion alarm, activation should automatically display the associated camera and the door view should provide Identification detail.

So the security function is not simply:

“There is a camera outside the NVR room.”

It is:

See the area → identify who is at the door → automatically present the camera when the alarm activates.

That is a much stronger operational requirement.

Common CCTV coverage mistakes

“The camera cone touches the doorway, so the doorway is covered.”
Geometric coverage does not prove that Identification or any other required level is achieved.

“We use 4K cameras, so all areas are Identification.”
Resolution alone does not determine target detail.

“The whole lobby needs Identification.”
The Standard generally assigns Observation to common areas and Identification to the frontal main entrance/exit view.

“If we can recognise a person, that is Identification.”
The Standard treats Recognition and Identification as separate surveillance objectives.

“120% is the Identification requirement for every modern camera.”
No. That is the traditional reference. The Standard’s digital conversion table gives 45% for 1080p as a guideline.

“45% at 1080p guarantees Identification.”
No. The Standard specifically says factors such as lighting and camera angle still affect image quality.

“The live image looks good, so the requirement is proven.”
Recorded quality needs to be examined because compression may reduce detail.

“A wide-angle camera is always better because it covers more.”
Greater scene coverage may reduce target detail.

“Emergency exits only need cameras on one side.”
The Standard calls for frontal views in both directions.

Practical CCTV Coverage Review Checklist

For every surveillance area, ask:

1. What is the security purpose of the camera?
2. What target are we interested in?
3. Where will that target be when the useful image is required?
4. Is the required level Detect, Observe, Recognise or Identify?
5. Does the Standard assign a particular level to this type of location?
6. Is the target expected to be frontal, side-on or viewed from above?
7. Does the selected field of view provide enough target detail?
8. Is the camera attempting to cover too much area?
9. Does the camera resolution support the target field of view?
10. What is the lowest resolution within the imaging and recording chain?
11. Could compression reduce the usable detail?
12. Will lighting affect the target at the required location?
13. Does WDR need to handle backlighting?
14. Is the camera angle suitable for facial detail?
15. Will people, queues or crowds obstruct one another?
16. Could signage, landscaping or building features obstruct the view?
17. Is general Observation coverage needed in addition to a targeted Identification view?
18. Are vehicle number plates a separate requirement?
19. Are emergency exits covered in both directions?
20. Are lift and stair access routes adequately covered?
21. Are Sensitive Rooms and Critical Areas covered according to their required level?
22. Does an intrusion alarm automatically show the relevant camera where required?
23. Has the installed field of view been documented?
24. Has the recorded image been tested, not just the live view?
25. Will this same requirement still be achieved after dark or under expected operating conditions?

Conclusion

The SPF VSS Standard changes the question from:

“Can the camera see this area?”

to:

“Can the camera see this area with enough detail for the intended security purpose?”

That is a much more useful way to design CCTV.

The four levels provide a clear progression:

Detect: Is someone there?
Observe: What are they doing?
Recognise: Is this someone we have seen before?
Identify: Who is this person?

The Standard then applies these levels according to the function of the building area.

General common areas normally require Observation.

Security or registration counters require Recognition.

Main building entrances require a frontal Identification view.

Sensitive and Critical Area access can require a combination of Observation and Identification.

For 1080p systems, the Standard’s digital conversion guidance gives approximately:

Detect 10% → Observe 10% → Recognise 20% → Identify 45% of image height, while explicitly warning that lighting, angle and compression must still be considered.

The most important practical lesson is:

Do not prove CCTV performance by showing that an area falls inside a camera cone. Prove that the required target reaches the required level of detail at the exact location where the security outcome is needed.