Every alarm on a hold baggage machine has to be resolved before the bag flies, and ours get resolved by opening them. The object the image shows has never once been in there, and in fourteen months of archived images it sits in the same place in the frame.
I screen hold baggage. Twelve years, trained to the level where I clear images rather than only look at them.
How it works, stripped of the parts I am not allowed to tell you.
The bag goes through a machine that measures how much of the beam gets through it and at what energies. From that it works out density and an approximation of atomic number, and it colours the image accordingly. Organic material comes out orange, metal blue, and the denser something is the darker it renders.
The machine decides on its own whether to alarm. If it does, the bag goes to a higher level of screening, and at the top of that ladder the bag is opened in front of two people and searched.
Every alarm is logged and every image is archived with the bag tag number and the time. That archive exists so that somebody can come back in two years and audit what I did.
About once a month the machine alarms on the same thing.
An object in the orange range, dense, about two hundred millimetres long, with a shape the classifier puts in the organic group.
We open the bag. It is clothes, shoes, a wash bag, a phone charger. The ordinary contents of an ordinary holiday.
There is nothing in it that is two hundred millimetres and dense and organic.
We re-run the bag through the same machine, which is procedure.
The image is clean.
What I have checked, and some of it with the auditor sitting next to me.
The daily test piece. Every machine in that hall is tested at the start of every shift against a manufactured sample with known objects in it, and the test is logged, and ours has never failed one.
Artefacts. This is the real answer and I went at it hard, because an X-ray image is full of them and an operator who does not know them is useless. A dense object casts a streak. A sharp edge rings. A metal hinge throws a shadow that can read as a second object. They all have a look, and they all move when the bag moves.
The auditor looked at nine of these with me in 2023. She agreed that it is not in the artefact families she knows, and she wrote that, and then she wrote that the machine should be serviced.
Two machines. There are two in that hall, a different generation from each other, and both of them do it.
The thing I found when I went into the archive.
I pulled fourteen months of them, twelve images, because I wanted to look at them side by side and the archive lets you do that.
The object is in the same place in every one.
Not the same place in the bag. Different bags, different sizes, packed by different people, oriented however the belt took them.
The same place in the frame.
Fourteen pixels in from the left hand edge, a little below the middle, lying along the direction of travel.
I know what that sounds like. A fixed position in the frame is the signature of a fault in the detector array, and that would be the end of it, and I would be relieved.
The two machines do not produce the same size of image. Different arrays, different resolutions, different manufacturers.
Fourteen pixels from the left on both.
What happens to those bags.
They fly. That is the correct outcome and I want to be clear that I stand behind every one of them. The alarm was resolved the way the standard says to resolve it: the bag was opened, searched by hand by two trained people, and found to contain nothing of interest.
I have never cleared an image I was not happy with and I have never been asked to.
I logged it as a machine fault the first four times. Then I started logging it as an unresolved image classification, which is a box nobody has ever asked me about, and the service engineer has been told about it three times by three different people and has found the machines within specification three times.
The last thing, which is mine and which I have not put on a form.
The bag tags.
Each of those twelve images has a tag number, and a tag number tells you the flight, and the flight tells you the check in desk and the time, because the system holds all of it and I am allowed to look.
Twelve different passengers. Eleven different flights. Four airlines, three countries, and a spread of people that has nothing in it: a family, two men travelling alone, somebody's grandmother.
All twelve bags were checked in at desk 12, between five and six in the morning.
I work the early shift. Desk 12 is the one you see from the screening room through the glass, if you stand up.