Originally Posted by
WillCAD
Your points are all valid, and I respect your decision to be abusively rubbed down rather than peeped. But I choose the AIT, even without the ATR software.
I will not be touched by government clerks! I don't care if it means I can't fly, if I'm stranded on the other side of the country and have to rent a car and drive 4 days to get home and miss work and incur huge additional costs. I will not allow them to put their hands on my person, for any reason, EVER. This is MY choice to make, and I don't really care whether others respect my choice or not.
You're using logical thinking, Ink, which doesn't apply terribly well to either TSA or law enforcement.
The info in this thread is the first I've heard about the possibility that ATR-equipped scanners don't actually compose an image from their scan returns but simply feed the raw returns directly to the ATR software. I had always assumed that the machine continued to perform its primary function of creating an image from the scan returns, and funneled either the image or the image plus the raw scan returns, to the ATR for pattern recognition.
Logical thinking and the TSA...I should have caught that.
In machinery analysis systems, technology with which I am familiar, they originally tried to analyze the data in the same manner as human technicians. It would complete all of the mathematical transforms and create virtual data representation similar to what would be "seen" by analyst. The analyst would provide feedback to the system and the system became smarter and smarter. Eventually, the system was able to "see" the patterns within the data itself without a human interpretable representation having ever been made.
This creates a serious problem. Problems that are unique and outside the previously seen problem set must be flagged for human analysis or ignored. It depends on how the thought process of the designer works. If they think "We know all that we can possibly know" then any new anomaly will be considered just that. It will be discarded as every known (read "possible") problem will already be in the system and anything new can not be a problem.
The second is equally problematic: "We will always be discovering new problems that will need to be programmed into the system." This requires a continuing analyst presence to determine the source and severity whenever new patterns present themselves. This continues the perfection of the detection/discernment/determination loop. This would be done by the ATR flagging an area, resolution of the anomaly by physical inspection and informing the system that the pattern is innocuous or should be flagged in the future.
We do not know which thought process is used. I suspect that all changes are made in the laboratory and software upgrades are sent to the system as improvements are made.