Pandoras box
Article Type: Editorial From: Sensor Review, Volume 31, Issue 2
The theme of this issue is “Image processing and pattern recognition”. This whole area is of particular personal interest to me because it was the basis of my postgraduate research. I am probably biased, but I consider that this field has probably developed further and faster than most others in the 35+years since I was messing around with 32×32 binary images.
To be honest it had to. It seemed interesting at the time, and we did manage a few pretty elementary successes like getting a Puma 260 robot to pick up dice from a backlit table and throw them again, but we did not even attempt to read if we had a pair of sixes.
The 32×32 camera was pretty small and would not look out of place size-wise even today, but if we wanted to go up a level in sophistication we were talking vidicon tubes and a swarm of high-speed RAM chips and a jolly expensive A/D converter to give us a 256×256 image with 256 levels of grey in a very 8-bit world – and a power supply the size of a shoebox and a noisy fan trying to prevent meltdown. But such effort was hardly worthwhile because the magnificent amount of data it produced took us too long to process for us to be able to do anything very useful with it.
Typing the above paragraph, I discovered that my OpenOffice word processor thinks “vidicon” is a typing error because it does not appear in its dictionary! It should count itself lucky I was not using plumbicons!
These days we have colour cameras with 10,000 times the resolution included almost free of charge in our mobile phones. And to add insult to our pained recollection of Stone Age struggles, the camera can tell we are not smiling and refuse to take a picture – or even worse, can image-process our faces to make us smile.
I actually find it very gratifying to see the staggering progress that has been made, especially in the last five years. Probably more by luck than good judgement, we saw an inkling of the potential of machine vision but were then frustrated by the magnitude of the task and the limitation of our processing power.
My local regional airport has a short-stay car park and when I press the button to raise the barrier the ticket has my car’s registration number on it. Not a photograph of my number plate, but a pattern recognition analysis of the number itself.
Google is mapping our streets and cities and our houses will soon be on public display. The immigration queue at the airport we are photographed and no doubt cross-referenced to a visual database of undesirables.
Where is this all leading? When I was working on primitive pattern recognition the main motivation (apart from academic advancement), was to solve a problem, to reject defective parts, to improve quality and to make the world a better place.
The trouble with developing new tools is that sometimes they prove themselves useful for originally unconsidered tasks. The same software that recognises terrorists will soon, and probably very soon, be watching our every move. Our privacy will be lost forever for a mixture of commercial and public welfare benefits.
Pandora’s box has been opened.
Clive Loughlin
