Logistics · packaging line

Conveyor object counting — the same count at 2, 5 and 10 frames per second

A camera counts the boxes passing on a belt and reads the label on each one. Three inspection jobs run the same conveyor video side by side at 2, 5 and 10 frames per second and count the 8 boxes in the same order. The run screen shows each job's counts, the last lap and the recent crossings with ID and label.

Example setup

This is a fictional packaging line, not a real site or customer installation. The camera video, boxes and labels are all computer-generated. It is a demo sample; the counts on the screen are not field-validated counting values.

3cameras — 2, 5 and 10 fps, same video
8boxes per lap (labels P205 … T172), 38.0 s
Samelast-lap count on all three jobs — 8 each
1run monitor (one 1280×720 landscape screen)

Running it

One lap: three jobs count from 1 to 8 side by side

The line is computer-rendered, and the monitor shows a recording of the run screen with the same sample running in LinkVion. The monitor text is small, so the 10 fps card (counts, laps and recent crossings) is shown close up at the top left, cropped from the recording at its original size.

Example setup

One lap (38.0 s): each time a box on the belt passes the yellow marks under the camera, this lap rises by one from 1 to 8 and the box adds its ID and label to the recent crossings. The counts in all three cards rise together in the same order · 38 s

The line and the boxes on the belt are computer-rendered for illustration. The belt boxes match the boxes in the monitor's synthetic video in colour, label and order, and are timed to pass the yellow marks under the camera at the count times logged during recording. You cannot run it on this page.

On the screen

How it counts

  • Three vision jobs
  • Motion regions and object tracking
  • Counting line (A→B, B→A)
  • Label reading per object
  • Folder replay cameras
  • Script collects crossings and laps
  • Korean and English screens
One video, three speeds
Three folder replay cameras play the same scene at 2, 5 and 10 frames per second. Frames × interval is 38.0 s for all three, so the scenes stay in step however long they run.
Motion regions and tracking
The empty belt is learned as the background first, and regions that differ from it become objects. Tracking links each region across frames under one ID and counts it as A→B or B→A when it crosses the counting line in the middle.
Label reading
For each newly confirmed object the label (one letter and three digits) is read from its region and attached to the event, up to 2 per frame and 3 tries per object.
Laps
A script gathers the last 5 crossings and the median per-frame time for each job. A gap of more than 8 s between crossings starts a new lap, and this lap's count moves to the last lap. The last-lap figure is where the three jobs are compared.

Limits

What this case does not show

  • Not a real site or real camera. The line and belt boxes are computer-rendered, with a recording of the same sample running in the product laid over the monitor. The rendered boxes are timed to the count, not objects the camera actually saw.
  • The video, boxes and labels are synthetic, and the motion, tracking and reading settings are tuned to this video. Do not use them as they are for real counting.
  • 2, 5 and 10 frames per second are values chosen by the sample, not processing limits.
  • The cumulative A→B count keeps growing when the video loops back to the start; only one lap was recorded.
  • There is no live version on this page, only the video. The sample needs the vision inspection feature to run.