Asset protection surveillance is coverage design for the things that move: what must be identified, from which angle, at what distance, and for how long the recording is kept. Yards are the hard case, because distance and darkness decide whether the video proves anything.
Theft from a yard, warehouse or laydown area rarely happens in one place. Materials move by vehicle, through a gate, at a time when nobody is watching the right camera. Surveillance that protects assets has to follow that movement and hand over evidence that still stands up months later.
We start from the inventory and the routes it takes out of the site, then design coverage that identifies at the points where identification matters.
Based in Westminster. Serving Denver and clients across Colorado. See all Colorado service areas.
| Goal at the subject | Approximate pixels per foot | Typical use |
|---|---|---|
| Detect that something moved | About 8 | Open yard, fence line |
| Observe what a person is doing | About 19 | Lane, gate approach |
| Recognize a known person | About 38 | Entrance, parking area |
| Identify a stranger or a plate | About 76 | Gate, door, cash point |
Start with what moves and where
Before a camera count, list the assets, their value density and how they leave. Copper, fuel, tools, catalytic converters, finished goods on pallets and equipment on trailers each leave differently. A high-value cage needs a different view from an open gravel yard.
- Assets ranked by value and by how easily they can leave the site.
- The routes out: gates, fence lines, rail sidings, loading doors and fire exits.
- The points where a person becomes a vehicle, or a vehicle reaches public road.
- Where the evidence has to end up: insurer, HR, law enforcement or a civil claim.
Coverage geometry and pixel density
A wide lens at a doorway identifies a person. The same lens at thirty meters identifies a shape. The useful measure is pixels across the subject, not megapixels on the box. Video surveillance standards commonly put identification at around 250 pixels per meter across the subject, which is roughly 76 pixels per foot; recognition of a known person works at about half that, and detection at a third of it again.
That number turns a vague request into geometry. At four meters a camera covers a door width; at thirty meters it covers a yard. Where the subject distance is long, a longer lens, a narrower view or a closer mounting point is the answer, not a higher-resolution camera in the same position.
Thermal and low-light options
Thermal cameras do not see color and cannot read a plate. They see heat, which makes them effective at detecting a person across a dark yard, in light fog or behind light vegetation. That makes thermal a detection tool, paired with a visible camera that can then be moved to the same point.
Low-light performance on a standard camera depends on sensor size, lens aperture and how much light reaches the scene. A camera marketed for low light, in a yard with no lighting, is a specification problem rather than a firmware setting.
- Thermal detection where lighting the perimeter is impractical.
- A visible PTZ linked to thermal detection, so the operator gets a usable image.
- Infrared range matched to the standoff distance, not to the marketing figure.
- Housing heaters where frost and condensation defeat the image.
Analytics that survive a real yard
Analytics work when the scene is constrained and the rule matches the risk. A line crossing across a gate approach at two in the morning is a usable rule. Motion detection across a yard with blowing debris, cats and headlights is not, and it trains people to ignore alerts.
Design rules around the asset, then tune them after real activity. Vehicle-classified alerts in a gate lane, object-left-behind on a loading dock and loitering near a fuel island are rules with a clear response attached.
- Object classification, so a person alert means a person.
- Rules placed on boundaries that matter rather than on every camera.
- A named recipient for each alert type, and a test that it arrives.
- A tuning period after handover, using real events rather than guesses.
Evidence that holds up later
An image that identifies someone is only useful if you can show when it was taken, from which camera, and that it has not been altered. That means a time source the recorder trusts, exports that carry their metadata, a checksum where the platform supports one, and a retention period long enough for the discovery timeline of a typical incident.
Access to footage matters too. If anyone can delete a recording, the footage is weak evidence. Role-based access, a log of exports and views, and a written policy on who may delete are as much part of the system as the cameras.
- Time synchronized across cameras and recorders, checked at handover.
- Export with metadata and a checksum where the platform provides one.
- Retention aligned to insurer and legal expectations, written down.
- View and export history recorded and restricted by role.
Connecting coverage to operations
Asset protection improves when the video feeds something other than security. A gate camera that captures plate reads, a dock camera used to check load condition, or a camera on a scale that records what actually left the site all produce a record operations wants as much as security does.
That is where a surveillance project starts to pay for itself: the same camera serves two departments, and the retention policy has to satisfy both of them.
Frequently asked questions
How much footage do we need to keep?
Long enough to cover the delay between an incident and its discovery, which for many sites is a month rather than a week. We size storage to the retention you choose and tell you what changing that figure would cost in disk and recorder capacity.
Does thermal replace a normal camera?
No. Thermal detects a person across a dark yard but shows no color, no plate and no detail. The pairing that works is thermal for detection and a visible camera, usually a PTZ, for identification once the detection is confirmed.
Can analytics watch a high-value cage?
Yes, and a small defined area is where analytics behave most predictably. Rules for entry, for a person remaining too long and for an object left behind are reliable in a cage or a dock, and unreliable in an open yard with traffic in the background.
Will video hold up in an investigation?
It will if the recording can be tied to a clock the recorder can prove, exported with its metadata and shown to be access-controlled. We check those three things at commissioning, because an image without them is hard to rely on later.