Companies use them in conveyor systems quite extensively. They can be used to detect anomalies, differentiate between different types of product, etc. the technology already exists, it would just need to be adapted and programmed to this specific use. Airports already use similar technology for baggage routing, although they're scanning tags instead of image processing.

They can only really include sensors for weight and stuckage without an expensive retrofit. You can have something that could potentially stop and inappropriate object, probably use it x-rays and computer recognition (or some guy at a monitor) but you see all of that costs money. They don't want to spend that money. Especially on such a fringe case. And let's be honest, nobody wants to spend millions of dollars trying to protect the dumbest possible people from themselves. I mean, they say you can't put a value on human life, but we all know that everybody does anyway. Especially when they're rich and in a position of any kind of authority.

I see many comments discrediting this somehow, but I want to put my two cents in as someone who does work with sensor based AI assisted processing in real time and safety reliant environments.

Just because a concept can be thought of that sounds reasonable and maybe even works in simple tests, that doesn't mean that it's actually useful for the real use case. Many typical approaches to creating models that can solve computer vision tasks such as this can result in unstable results and no system that has a considerable false positive rate would be tolerated by any airliner. This isn't even to speak of the false negative rate which might then still be rather high, which still leaves the system useless.

Naturally it's not to say that no such system could be created, but they can't be just whipped out like some people here claim. If, as people here are already assuming, the problem happened because someone climbed onto the conveyor belt and was carried in, then this type of problem is sufficiently unthinkably rare that most companies didn't think about it much either.

Clearly greater security is necessary, but people are being unreasonable with how trivial they portray the solution as being.

I mean. It’s not magic. Even I could run a basic computer vision system from a raspberry pi and a webcam. They could easily come up with a system that returns confidence percentages for bags vs humans. Assuming the baggage isn’t coming in too warm, they could have an IR thermal camera that stops the line if it detects any temperature over a certain threshold. They could also use other tests, like two pictures a few seconds apart to see if something is moving on its own on the belt. I’m sure there are even more tests than the ones above, let alone design changes that could disincentivize folks doing dumb stuff, or making their dumb stuff easier to spot - like putting the end of the return conveyer behind plexiglass, so the person is visible on the conveyer before they disappear into the wall.

All of the above are not 100% solutions, but taken together, they can establish a reasonable confidence level they’re not about to intake something that isn’t a bag.