By Luis Lopez, AI transportation consultant, CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast
Every container that enters or leaves a terminal, depot or warehouse yard passes through a gate. At that moment, someone has to confirm which box it is, which chassis it rides on, which truck is pulling it, whether the seal is intact and whether the equipment arrived damaged. For years that meant a clerk with a clipboard or a driver reading numbers into an intercom. Today, more gates use cameras and AI to capture that information automatically.
This article explains how gate camera systems read container numbers, plates and seals, how they connect those reads to appointments and bookings, and where they still need human help.
What a Gate Camera System Actually Does
A typical setup places several cameras around a gate lane: overhead, on both sides, front and rear. As a truck rolls through, the cameras capture a sequence of images. Software then:
- Finds the container number painted on the box and reads it with optical character recognition
- Reads the chassis number and the truck’s license plate
- Captures the seal and, where possible, the seal number
- Photographs all sides of the container and chassis for damage records
- Matches everything against appointments, bookings and release records
The output is a gate transaction record with images attached. If everything matches, the truck can proceed. If something does not, the transaction goes to an exceptions queue. For a broader look at vision tools in yards and warehouses, see computer vision in warehouse, yard and freight operations.
Reading Container Numbers With the ISO 6346 Check Digit
Container numbers follow the ISO 6346 standard. Each number has a three-letter owner code, an equipment category identifier (U for freight containers), a six-digit serial number and a single check digit. That structure is a gift to OCR.
The check digit is calculated from the other characters. When the camera reads a number, the software can recompute the check digit and compare it to what it read. If they do not match, the system knows at least one character is wrong. Good systems then:
- Compare reads from multiple cameras and pick the one that validates
- Try likely character substitutions, such as a zero misread as the letter O, and test whether the corrected number validates
- Check the candidate number against expected containers for that appointment
- Send the read to a human if no candidate validates with confidence
A valid check digit does not prove the read is correct, since a different valid number could exist, but it removes a large share of obvious errors before they enter the system.
Chassis and License Plate Reads
Chassis numbers and truck plates are harder. Chassis markings vary in location, size and condition. Plates vary by state and can be dirty, bent or partly hidden by a hitch or a bumper. AI models trained on many examples handle these variations better than older rule-based OCR, but accuracy is still lower than for container numbers, partly because there is no universal check digit to validate the read.
Matching helps. If the system knows which trucking company has an appointment and which plates that company registered, it can confirm a partial read instead of guessing.
Seal Number Capture
Seals protect cargo integrity, and seal numbers appear on shipping documents. Cameras can confirm that a seal is present on the door and, with the right angle and resolution, read the seal number. Seal numbers are small, sometimes embossed rather than printed, and often partly covered by the seal’s own shape, so many operations still rely on a handheld scan or manual entry as a backup.
Even when the number cannot be read, a clear photo of the seal at the gate is valuable evidence. Our guide to trailer and container seals explains why the seal record matters when questions arise later.
Damage Photos at In-Gate and Out-Gate
Gate cameras create a time-stamped visual record of the container and chassis condition at each handoff. When a dent, hole or broken door hardware is found later, comparing in-gate and out-gate images helps establish when the damage appeared.
Some systems now use AI to flag likely damage automatically, such as holes, large dents or missing parts. Those flags should be reviewed by a person before anyone is billed. See AI trailer damage inspection from photos for more on how these models work and where they fall short.
Matching Reads to Appointments and Bookings
A read on its own is just data. Its value comes from matching. At an in-gate for an export, the system checks the container against the booking. At an out-gate for an import, it checks whether the box has been released and whether the trucker holds a valid appointment. At a warehouse, it checks the container against the expected receipt.
When all of these match, the gate transaction can be nearly hands-free. When they do not, the system should say exactly what failed: wrong container for the booking, no appointment, plate not registered to the trucking company, or seal number different from the document. Drayage teams that understand the appointment side will find port terminal appointments for drayage a useful companion read.
The Exceptions Queue
No camera system reads everything. The exceptions queue is where the operation either works smoothly or backs up. A good queue:
- Shows the clerk all images for the transaction side by side
- Highlights the field that failed and the system’s best guesses
- Lets the clerk correct a read with one or two actions
- Records the correction so the model can be retrained
- Escalates mismatches that suggest a wrong container or unauthorized pickup
Staff the queue for peak hours, not average hours. A slow queue turns a modern gate into a longer line.
Weather, Dirt and Angle Limits
Cameras are only as good as what they can see. Common problems include:
- Rain, glare and low sun angles washing out characters
- Night lighting that creates shadows or reflections
- Dirty, faded or repainted container markings
- Trucks entering at an angle or too fast for clean frames
- Stickers, placards or damage covering part of a number
Lane design, lighting, camera cleaning schedules and speed controls matter as much as the AI model itself.
Privacy and Plate Reading
Reading license plates and photographing drivers raises privacy questions. Some states have laws on automated plate reading, data retention and data sharing. Before installing plate cameras, check the rules where the facility operates, set a clear retention policy, limit who can access the images and post appropriate notices. Use the data for gate operations, not unrelated purposes.
What to Measure
Track a few clear measures to know whether the system is working:
- Share of transactions completed without manual correction, broken out by field
- Time per truck through the gate lane
- Exceptions queue size and time to resolve at peak hours
- Number of mismatches that turned out to be real problems
- Damage disputes where gate images settled the question
As an AI transportation consultant, the first thing I look at at a camera gate is the exceptions queue, because that is where hidden labor and delays live. A gate that reads well in good weather but clogs every rainy afternoon is not finished yet.
For more conversations on freight and AI, subscribe to the Freight Guru Podcast.
About the author: Luis Lopez is a Miami-based AI transportation consultant and logistics entrepreneur, the CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast.
