Computer vision · Container codes
The code on the containergoes straight into the system.
Container code recognition is the optical character recognition (OCR) of the ISO 6346 identification code painted on a container, that is the owner prefix, serial number and check digit, directly from camera images. We build systems for gates, terminals and depots that read the code, check it against the rules of the standard and send it to the terminal operating system, with a person confirming unclear cases.
Events
- 09:15:27Truck at gate 2 · 40' container
- 09:15:28Code read · LMDU 012345 1
- 09:15:28Check digit is valid
- 09:15:29Sent to the terminal system
A mistyped code costs time at the gate
At many gates, the container code is still read by eye and typed in by an operator, in the rain, at night or with a lorry waiting. A single wrong letter means a container "lost" in the system and hours spent searching the yard.
- Codes are transcribed by hand, with one-letter or one-digit mistakes that are hard to spot
- Lorries wait at the gate while the operator records and checks the details
- There is no systematic photo of the container on arrival, useful in disputes about its condition
- Gate data reaches the TOS or ERP late or only after being entered a second time
Who it is for
- Container terminals and ports, including operators around the port of Constanța
- Empty and full container depots that record gate-ins and gate-outs by hand
- Rail operators and haulage companies that hand over or collect containers at the gate
- IT teams that want to feed the TOS or ERP with automatically read codes
Step by step
How the system reads a container code, step by step
The five steps on the animation follow a truck through the gate: the image, finding the code, reading it, validating it against ISO 6346 and the person who confirms unclear cases.
Events
Events
- 09:15:27Truck at gate 2 · 40' container
Events
- 09:15:27Truck at gate 2 · 40' container
- 09:15:28Code read · LMDU 012345 1
Events
- 09:15:27Truck at gate 2 · 40' container
- 09:15:28Code read · LMDU 012345 1
- 09:15:28Check digit is valid
- 09:15:29Sent to the terminal system
Events
- 09:15:27Truck at gate 2 · 40' container
- 09:15:28Code read · LMDU 012345 1
- 09:15:28Check digit is valid
- 09:15:29Sent to the terminal system
Step 1 of 5
The camera at the gate
Cameras are mounted at the side, at the rear and, where possible, above the lane, at the height of the painted codes. Capture starts when a presence sensor or the barrier signals the vehicle, and at night dedicated lighting keeps the characters visible. A frame is good when the letters are clear, upright and free of strong reflections.
Where the reading is installed
Portal at the truck gate
A frame with cameras at the sides, rear and top reads the container while the truck passes slowly. This is the most common option at terminals and depots, for recording gate-ins and gate-outs.
Cameras on the crane
Cameras mounted on the crane read the code during lifting or unloading. It suits quay and yard operations, where the container does not pass through a gate.
Rail portal
Trackside cameras read the codes of the containers on the wagons one after another, and the list is compared with the shipping documents. It needs careful set-up of camera angles and of the train's passing speed.
What you receive at hand-over
- The model that finds the code area and the OCR model, adapted to the containers at your site and exported for the gate hardware
- Camera positions, trigger signal and reading zones configured for each checkpoint
- ISO 6346 validation rules and rules for combining reads from several frames and cameras
- Integration with the TOS, ERP or gate application, typically via API, EDI or CSV files
- The review screen for operators, with the photos of every flagged passage
- Documentation, operator training and a measurement report on real passages at your gates
No strings attached: tell us what the cameras should see and we'll tell you honestly whether it can be done.
Let's talkWhat the recognition system can do
- 01
Finds the code on the container
The model locates the code area on the doors, the side or the roof, including codes written vertically.
- 02
Reads the prefix, serial and check digit
OCR extracts the three-letter owner prefix, the category identifier, the six-digit serial number and the check digit.
- 03
Validates the check digit
The check digit is recalculated using the ISO 6346 algorithm; a code that does not match is sent for review automatically.
- 04
Reads the size and type code
Where it is visible, the four-character code describing the container's length, height and type is read as well.
- 05
Combines readings from several cameras
Results from the side, rear and top cameras are compared, and the system keeps the most reliable reading.
- 06
Sends the data to your systems
The code, time, gate and photos are sent via API or messages to the TOS, ERP or gate application.
How it works
How a container code gets from the image into the system
The steps marked in yellow are approved by a person.
- Input
- AI
- Action
- Human in the loop
- Result
- 01
Input
Gate cameras
Cameras capture the container from the side, the rear and, where possible, from above.
- 02
AI
Location and OCR
The model finds the code area and reads the characters.
- 03
Action
ISO 6346 validation
The check digit and the format are checked against the rules of the standard.
- 04
Human in the loop
Operator confirmation
Uncertain or invalid readings are confirmed or corrected by a person.
- 05
Action
Sending to the TOS
The validated code and photos reach the terminal system or the ERP.
- 06
Result
Recorded gate move
You have a record of the container, with time, gate and images.
What container code recognition needs
- Cameras mounted at the gate or on the crane, with an angle and resolution suited to the painted characters
- Additional lighting for night-time, where existing light does not reach
- Real images from the site with containers from several owners, including dirty, faded or partly covered codes
- A signal that marks the vehicle passing, for example a presence sensor or the gate barrier
- Access to the terminal operating system or the ERP, via API or exchange files
- Processing resources at the gate or on a local server, so that reading does not depend on the internet connection
Possible integrations
- IP cameras with an RTSP stream
- ONVIF-compatible cameras
- Terminal operating systems (TOS)
- ERP systems, for example SAP
- Gate applications and access control systems
- Barriers and presence sensors
- NVIDIA Jetson edge devices
- REST API and webhooks
- EDI messages or CSV files
- PostgreSQL databases
- IP cameras with an RTSP stream
- ONVIF-compatible cameras
- Terminal operating systems (TOS)
- ERP systems, for example SAP
- Gate applications and access control systems
- Barriers and presence sensors
- NVIDIA Jetson edge devices
- REST API and webhooks
- EDI messages or CSV files
- PostgreSQL databases
Each integration's compatibility is checked during the analysis phase, for the systems and versions you use.
Control and transparency
What stays with people
- Confirming low-confidence readings or readings with an invalid check digit
- Deciding whether to let a container in or out when the data does not match the booking
- Checking the container's condition; photos help but do not replace an inspection by a person
- Setting the image retention period, especially when licence plates or people also appear in the frame
Limits and operating conditions
- Dirty, rusty, repainted codes or codes covered by labels may be misread or not read at all
- The check digit catches many reading errors, but not all of them; a valid code may still belong to another container
- The angle, vehicle speed, rain, reflections and poor light directly affect reading quality
- Integration with the TOS or ERP depends on the interfaces their vendor provides and is checked for each project
How we evaluate container code recognition
We agree the indicators together before launch and track them afterwards. The values depend on the process, the data and the volume.
- Fully correct readings
- The share of codes read correctly character by character, compared with a sample checked manually by operators
- Codes sent for review
- How many passages reach the operator because of low confidence or the check digit, tracked daily
- Errors that pass validation
- Codes accepted automatically but later corrected in the TOS or ERP, counted and analysed by cause
- Results across conditions
- Reading quality compared separately by day and night, in dry weather and rain, on clean and worn containers
How we build container code recognition, in 4 steps
- 01
We assess the gate and the flow
We look at how vehicles pass, where cameras can be mounted and what data the terminal system needs.
- 02
We collect and label images
We record real passages and label the codes, including hard-to-read ones, so that the model learns them.
- 03
We train, validate and test
We adapt the OCR model, add ISO 6346 validation and test on passages the system has not seen.
- 04
We integrate and monitor
We connect the results to the TOS or ERP and track operator corrections, which feed retraining.
Use cases
The scenarios below are conceptual examples, designed to illustrate realistic applications. They are not implementations delivered by WEB-AZ.
- Port terminal
The lorry gate
At the terminal entrance, the container code is read and validated, and the gate operator only sees the passages where the system is not sure.
Conceptual example - Container depot
Tracking empty containers
Each container entering the depot gets a record with its code, time and photos from all sides, useful when it is handed over to the customer.
Conceptual example - Rail freight
Containers on wagons
A camera at the terminal exit reads the codes of containers on wagons, and the list is compared with the shipping documents.
Conceptual example
Useful guides
Before you decide, read
Frequently asked questions
What is automatic container code recognition?
It is the optical character recognition (OCR) of the identification code painted on a container, as defined by the ISO 6346 standard, directly from camera images. The code is checked against the rules of the standard and sent to the terminal or depot system, or to the ERP.
What does an ISO 6346 code contain?
The code has three letters for the owner, one letter for the equipment category (for example U for freight containers), a six-digit serial number and a check digit. Separately, the container also carries a four-character size and type code.
What is the check digit for?
The check digit is calculated from the other characters of the code using an algorithm defined in the standard. If the digit read does not match the calculated one, the system knows that at least one character is wrong and sends the reading to an operator.
Does it work with the cameras we already have at the gate?
Sometimes, but cameras mounted for general surveillance rarely have the right angle and resolution for the characters on a container. We review real images from your gate before confirming whether they are enough or whether dedicated cameras are needed.
Can licence plates be read at the same time?
Yes, reading the lorry's plate can be added at the same checkpoint, but licence plates are personal data. Using them requires a clear purpose, an agreed retention period and informing the people filmed, in line with the GDPR.
What happens when a code cannot be read?
The passage is not blocked automatically but flagged for review. The operator sees the photos, confirms or corrects the code, and the correction is later used to improve the model.
Can’t find the answer? Write to us and we’ll reply directly.
Want your gates to read container codes on their own?
Send us a few real images from the gate and tell us which system you use for records. We will tell you what can be automated and what should be checked on site.
Related solutions
- Licence plate recognitionAutomatic reading of licence plates at gates, barriers and car parks, with clear rules on access, retention and human review.
- Object detectionModels that recognise the objects that matter to your process in images and pass them on as structured data.
- Edge AI for computer visionVideo analysis on devices installed on site: fast response, less network load and images that never leave the building.