What is a video heat map and how is it made?
A heat map starts with object detection. A model finds people or vehicles in each frame, and a tracker links those detections across frames so one person is followed as one object. Each position is recorded against the image, usually as a grid of small cells. Over minutes, days or weeks, the system adds up how many objects, or how much time, fell into each cell. Longer collection periods smooth out one-off events and make the pattern more representative of normal use.
The totals are then converted into colors and drawn on top of a reference image from the camera. Areas with many detections appear in warm colors such as red and orange, areas with few in cool colors such as blue, and empty areas stay clear. The result is a picture of activity that compresses hours of video into one view. Because it relies on detection, the camera must have object analytics enabled for the map to exist.
What can a heat map tell you that video cannot?
Watching video shows individual events. A heat map shows habits. It can reveal the path most people take through a yard, the corner where vehicles queue, the doorway that gets the most traffic, or the area where people linger. Those patterns are almost impossible to see by scrubbing through recordings, because they only emerge when thousands of small moments are combined. That is the real value: it turns a pile of recordings into evidence about where attention, lighting or detection zones belong.
For security, that means finding where to focus. If the map shows repeated activity along a fence line at night, that is a candidate for a detection zone or a deterrent light. For operations, it can show congested loading areas or underused spaces. Compare maps from different periods, such as weekdays and weekends or day and night, to see how behavior changes. Treat the map as a prompt for questions, not as a verdict.
Dwell, flow and counts: which heat map is which?
Not every heat map measures the same thing. A traffic or presence map counts how many objects passed through each area, so it highlights popular routes. A dwell map measures how long objects stayed, so it highlights places where people stop, wait or loiter, even if few pass through. The two can look very different over the same scene, and mixing them up leads to wrong conclusions. Labeling each map clearly with its measure and period avoids arguments later about what the colors meant.
Some systems separate people from vehicles, or show direction of travel so you can see flow, not only density. Ask which measure a map uses and what time window it covers. A map over a single hour behaves differently from one over a month. People counting and loitering detection use the same underlying detection and tracking, so a platform that offers them can often derive maps from the same data.
What affects heat map accuracy and usefulness?
A heat map is only as good as the detections behind it. Camera placement matters: a high, angled view shows people clearly, while a low or grazing angle causes overlap and missed detections. Distance matters too, because small, distant figures are harder for models to find. Lighting, glare, rain and obstructions can all reduce what the model sees, and areas where it sees poorly will look artificially cool. Treat cool areas with suspicion until you have checked that the camera could actually see them.
Perspective also distorts the picture. Because the map is drawn on a camera image, a cell near the camera covers a small patch of ground and a cell far away covers a large one. That makes direct comparison between near and far areas tricky. Keep the camera fixed, since moving a PTZ changes the scene and resets the meaning of each cell. Verify the map against what you know happens on site before acting on it.
How should you use heat maps responsibly?
Start with a question, such as where do people cross the lot after hours, then choose the period and the camera that can answer it. Review the map alongside sample clips to confirm the pattern is real. Use findings to adjust zones, lighting, patrol routes or layout, and check later whether the pattern changed. Document what you changed so the next review is meaningful. Revisit the same view after changes to confirm that they had the effect you intended.
Heat maps usually show aggregate positions, not identities, which is a privacy advantage over reviewing individual footage. Even so, area use can still be sensitive, particularly where employees are involved. Tell people what analytics are in use, limit who can view maps, and confirm notice and data rules with counsel or the local authority. Do not use a map to judge individuals, since it describes places, not people. Retention of the underlying data should also follow a written policy.
Key terms
- Heat map
- A color overlay on a camera image showing where detected objects appeared most often or spent the most time.
- Object tracking
- Linking detections of the same person or vehicle across frames so movement and time spent can be measured.
- Dwell time
- How long a detected object stays in an area, used to find places where people wait, linger or loiter.
- Grid cell
- A small region of the image in which detections are counted; totals per cell become the colors of the map.
- Aggregate
- Combined data across many events and a long period, which shows patterns instead of individual incidents.
- Detection zone
- A drawn area of the image where analytics apply, often placed where a heat map shows repeated activity.
For a deployed trailer, a heat map can show where people and vehicles actually approach a perimeter over days, which helps decide where to aim fixed cameras or place a detection zone. AI VisionStream supports heat maps, with object detection enabled per camera. On a Gen 9 head, the three fixed cameras give stable views that suit mapping, while the PTZ moves and is less suited to it.
Common questions
Do heat maps identify specific people?
Generally no. They add up positions of detected objects over time, so they show where activity happens, not who was there. Privacy rules can still apply to monitoring areas, so confirm requirements with counsel or the local authority first.
Do I need special cameras for heat maps?
Not special hardware, but you need clear, stable views and object detection enabled on the camera. A fixed camera with a good angle and enough resolution on the subjects produces more reliable maps than a distant or moving view.
How long should a heat map period be?
Long enough to capture normal behavior. A single day may reflect a one-off event, while several weeks smooth out noise. Compare periods, such as weekday and weekend or day and night, to see which patterns are real and which are accidents of timing.