Kenny’s 3rd Law of Vision
7 February 2009
My 3rd Law of Vision isn’t related to my current task, so I thought I would logically discuss it first…. huh? Yeah, that’s how my mind works, sorry.
The 3rd law is called Kenny’s Eye Rule. The idea behind it is that vision system can get difficult or complicated when the object(s) being analyzed are moving faster than the eye can see. By being difficult, it means you may have to make special choices, play tricks, purchase expensive hardware, etc…
Many people don’t realize it but our eyes cannot see things that have a periodic frequency of not much greater than 50 Hertz (Hz) or objects that move back and forth 50 times per second. It’s true! The video we watch typically comes at us in discrete or digital frames. It’s true, even if it’s an analog TV. You can see this with any TV/computer, where you can can get temporal aliasing. You know, that effect when it looks like spokes on a wheel are spinning backwards. This is our natural humanity given built-in 50Hz threshold and it’s why TVs and monitors run at, or better than that frame rate.
The original analog TV system didn’t. Those systems actually run at a 30Hz frame rate (US), which would be really annoying if they didn’t pull out a trick on us. What those tricky engineers did was transmit half of the picture, the so-called even lines, and then the odd lines of the frame. These are often called the old and even fields, but what it gives us a 60Hz field rate, which spatially tricks our eyes to not notice, as much, the 30Hz frame rate that would be bothersome. Good trick, smart guys.
Why the history? Well the tools we tend to use in machine vision are based on standard video cameras and thus that 30Hz frame rate sepratated by two fields. There are special cameras, but as you might suspect, they are expen$ive and we try to avoid them, if we can. Hell, if a $50 camera does the job, use it! Sometimes they don’t, so…..
That’s where Kenny’s Eye Law comes in. We most often use those cameras and thus are bound by their limitations of frame and field rate.
I developed Rule of Eye, when I was helping a client develop a system that inspected the top and inside sides of bottle caps in Italy. Yes, the bottle caps that seal the top of beer bottles and the like. This was about 15 years ago and those machines were a mechanical wonder for me. The machine formed these bottle caps from flat metal and applied the inner seal at >50 caps per second. At that speed, the caps look like a continuous piece of metal. This system had two cameras: one inspecting the printer on the top and one inspecting the ‘crowns’ or desired number of folds in the metal and the quality of the inner seal.
So, that system worked faster than a standard video camera of 30 frames per second. How did we do it? We used the reverse trick of the original TV engineers, we just used 1/2 of the image and we used both the even and odd halves. That way way we could capture 60 images per second with a standard camera. Pretty cool, eh!
Another system, I worked on for a client more recently (ah um 8 years ago?), was a adaptive mirror control system. This company’s product was used in fancy pants telescopes. Its clients including thos looking up into space, often scientists, and those looking down from space to earth, often the military or spies. The goal of the adaptive optical systems was to adjust and remove as much of the optical distortion caused by our atmosphere and produce a clearer image. Wow! Don’t ask me about the physics of it or the optical system worked. But, I was able to help with the machine vision task they needed to do.
What we needed to do was track a number of dots where each represented a mirror position in XY or tilt or distortion in this case. The first and primary system I worked on had 37 mirrors or dots across the field of view, each moving slightly betwen the frames of the images. To do the system’s job, we had to track each of those dot positions and feed the XY cordinates into it’s closed loop control system. That control system would calculate and set an adjustment of the tilt for each of the mirrors with every new image and new XY positions of the dots. To effectively manage this control system, they needed to track these positions at almost 1000 image per second. A kiloframe per second! (think Doc in Back to the Future) Huh, we can’t do that can we? Yup, but we had to buy a very expensive camera that could capture images at the rate.
Needless to say, I hope! Having a system that moves faster than your eye can see complicates the development of a machine vision system. That’s called Kenny’s Eye Rule and my 3rd Law of Vision.