Posted: Wed Jun 08, 2005 4:46 pm
Not kidding at all… here is the story.
Robert had to carry out a project for the industrial experience year of his degree and I suggested that he made a system to compare similar objects and detect differences, showing the results on a TV picture. The solution was based on various modules that had been developed over the years and then he added some extra bits and the control software. Here is how it worked:
Point a monochrome camera at the object
Extract the line rate from the video signal.
Create an orthogonal sampling clock using a phase locked loop.
Digitise the video and write it into RAM… this become the reference image.
Continue to sample the output of the camera, displaying it on the RGB inputs to a standard RGB TV monitor, but limit the video range to about 80%
On a pixel by pixel basis compare the digital data still coming from the digitiser with the reference held in RAM
When the difference exceeds a chosen level top up the red channel to 100%... thus anything that moves gets identified, twice. Why?
In tests he found that pointing the camera at a clock showed a red edge on the minute hands after only a few seconds. This allowed him to check that the clock was working!
Happy days…
Robert had to carry out a project for the industrial experience year of his degree and I suggested that he made a system to compare similar objects and detect differences, showing the results on a TV picture. The solution was based on various modules that had been developed over the years and then he added some extra bits and the control software. Here is how it worked:
Point a monochrome camera at the object
Extract the line rate from the video signal.
Create an orthogonal sampling clock using a phase locked loop.
Digitise the video and write it into RAM… this become the reference image.
Continue to sample the output of the camera, displaying it on the RGB inputs to a standard RGB TV monitor, but limit the video range to about 80%
On a pixel by pixel basis compare the digital data still coming from the digitiser with the reference held in RAM
When the difference exceeds a chosen level top up the red channel to 100%... thus anything that moves gets identified, twice. Why?
In tests he found that pointing the camera at a clock showed a red edge on the minute hands after only a few seconds. This allowed him to check that the clock was working!
Happy days…
I just downed a glas of Laphroaig and am laughing my a** off.