Counting Line Guide

Drawing Counting Lines — The 3-Step Accuracy Check

Counting accuracy always comes back to one simple idea:

note

Can the system reliably track a person crossing a line from start to end?

Everything you do here is just a way to verify that visually before you save the line.

Step 1 — Check the Start–End Map (Is tracking healthy?)

Before drawing any line, always open the Start–End Map.

What you want to see:

• Most start points near the edges of the tracking zone
• Most end points near the edges of the tracking zone
• Very few random dots floating in the middle

What this tells you:

• People are being detected early
• Tracking is continuous
• Objects are not appearing or disappearing randomly

warning

• Many dots scattered in the middle → broken tracking
• Heavy “noise” areas → false detections or occlusion
• Start/end points not reaching edges → tracking zone too small or obstructed

If the start–end map looks messy, do not draw the line yet. Fix camera position or tracking zone first.

Step 2 — Review the Path Map (Where do people actually walk?)

Next, open the Path Map.

This shows the actual walking trails of people.

What you want to see:

• Clear, continuous paths
• Traffic flowing smoothly from one side of the zone to the other
• Minimal broken or zig-zag paths

Why this matters:

• The longer the tracked path, the more reliable the crossing detection
• Broken paths usually mean missed frames, occlusion, or poor positioning

Good rule of thumb:
If paths look clean and reach the edges, the system is ready to count.

Step 3 — Draw the Counting Line (Cut across the flow)

Now draw the counting line.

How to place it

• Draw the line across the main flow, not along it

• Place it where paths are dense and continuous

• Cut through the middle of the path flow

• Keep it away from:

  • Windows
  • Outside reflections
  • Plants, shadows, or irrelevant motion

Why this works

• The system tracks a person from edge → edge
• A clean line crossing confirms a valid count
• Short or broken paths reduce confidence

Once drawn, save the line.

That’s it.

How to Know You’re “Good Enough”

If all three visuals agree, accuracy is ready:

Visual Check	What You Should See
Start–End Map	Points mostly at edges
Path Map	Long, continuous paths
Noise Map	Minimal random dots

If you see:

• Excessive noise → false positives
• Broken paths → false negatives

Then the site is not ready yet.

What Comes Next (Optional Validation)

In the next section, you may:

• Generate a certification report
• Review scheduled validation videos
• Compare counts against ground truth

But those are confirmation tools.

The real accuracy decision is already made here — visually.

Supplementary 1 — Handling Hovering Staff & Security Guards

In many entrances, security guards or staff may stand near the doorway for long periods.

This is expected — and it is not a problem when handled correctly.

Why this matters

• A person who stays in view for a long time is not a real visitor
• Without filtering, hovering staff can inflate traffic
• Reliable tracking allows us to identify and ignore these cases safely

How the system handles it

• As long as the tracking zone fully covers the hovering area, the system can observe:

  • Entry time
  • Dwell duration
  • Movement pattern

• If a person remains in view longer than a defined time, they are automatically discounted

• They will not trigger counting lines, even if they move slightly

What you need to do

• Ensure the tracking zone includes the hovering area
• Enable hovering / dwell-time exclusion
• Set a reasonable time threshold (e.g. several seconds)

Result: guards and staff are tracked, recognised, and ignored — without affecting real traffic.

Supplementary 2 — U-Turns & Bounce-Back Movements

Sometimes people step inside, look around briefly, then turn back out.

This is common — especially with wide-angle cameras that see deeper into the store.

Why this matters

• Not every crossing should be counted as a true visit
• Very short in-and-out movements can distort accuracy

How the system handles it

• After crossing a counting line, the system waits briefly
• If the person U-turns within a short time, the crossing may be ignored
• Only meaningful entries are counted

This works best when:

• The tracking zone is large
• Paths are long and continuous
• The line is placed after the commitment point, not at the door edge

Result: quick bounce-backs are filtered, real visits remain accurate.

Supplementary 3 — Start & End Zones for Noisy Entrances (Advanced)

Some entrances are inherently noisy:

• Swing doors
• Glass reflections
• Tight vestibules
• Heavy in-store movement close to the entrance

In these cases, drawing a simple in-out line is not enough.

The challenge

• You must draw the counting line inside the store

• But that risks capturing:

  • In-store circulation
  • Broken paths near doors
  • Unreliable start/end detections

The solution: Start & End Zones

Start and end zones define where a valid track is allowed to begin or end.

How it works:

• People must start outside the zone and cross fully through

• Tracks that begin or end inside the zone are ignored

• A gap is kept between:

  • The start/end zone
  • The counting line

This ensures:

• People genuinely cross the entrance
• In-store movement is excluded
• Door noise is filtered out

When to use this

• High false positives near entrances
• Broken paths caused by doors or reflections
• In-out lines must be drawn deeper inside the shop

These setups can be very effective — but may require experience.
For difficult entrances, FootfallCam can assist with tuning.