Container yard with vision: how many there are, which ones, and in what condition.

Fixed cameras, or cameras on cranes and gates, that count containers, read their codes and log visible damage, with no walking the yard with a spreadsheet.

6 min readABrick Consulting
Stacked containers in a port yard with detections and decoded codes
Example image · codes read with OCR and validated against ISO 6346; boxes and condition hand-labelled
The essentials
  • 01Counts containers per block and row, and spots those that changed position.
  • 02Reads the container code and validates it with the ISO 6346 check digit.
  • 03Logs visible condition (rust, dents, doors) with photo and date.

The problem

In many yards, container inventory means walking the rows with a spreadsheet or a handheld. It is slow, out of date within hours, and typos in the code are common.

Computer vision turns gate cameras, cranes or a drone into a continuous inventory: which container is where, and what shape it arrived in.

How it works

Detect, read, validate and log the condition of every container.

Computer Vision · YOLO, OCR, ISO 6346
  1. 01 · DetectionThe model finds each container and counts it by block and tier.
  2. 02 · ReadingOCR reads prefix, number and type; the check digit rejects wrong reads.
  3. 03 · ConditionVisible damage is flagged with a photo, for inspection or claims.
Example simulation · real OCR + ISO 6346 validation; boxes and condition hand-labelled
In operations

Where it applies

  • Yard inventoryCounts by block, row and tier, continuously updated.
  • Gate in / gate outAutomatic code reading at gates and weighbridges.
  • Damage inspectionRust, dents and doors, with photographic evidence.
  • ReconciliationCross-check against the TOS or WMS to catch missing and misplaced units.
What it takes

The model is the easy part. The hard part is the operation.

Cameras
Gates, cranes or drone runs, with a view of the doors or sides.
Data
Images of your containers and typical damage to tune the model.
Integration
TOS, WMS or appointment system to match code, position and movement.
Operations
A review loop for doubtful reads and detected damage.
Technology

What we build it with

Model
  • Ultralytics YOLOObject detection
  • PaddleOCRCode reading
  • OpenCVImage processing
Infrastructure
  • NVIDIA JetsonEdge computing
Integration
  • PostgreSQLOperational data
  • GrafanaDashboards and monitoring

Reference tools. On each project we choose based on the architecture and systems the company already runs.

FAQ

The ISO 6346 check digit catches most wrong reads. Those are flagged for review instead of being loaded incorrectly.

It detects damage visible from the camera. The model is trained on examples from your operation and improves with use.

Yes. ISO 6346 is the same in every port and terminal, and the integration adapts to each yard’s system.

Does it make sense for your operation? Let's look at it together.

Tell us your case