Machine Vision

One of the things that Auto-Place started did was to begin using non-contact sensors for inspection. I believe my father had a fascination with lasers when they first came out and started on ideas to use them to inspect parts. I recall him telling me the story of about how the guy who invented the laser came up with the idea while sitting on a park bench. This all started in the mid 1970s and started to get a lot of recognition by the late 1970s.


We used lasers for applications to detect holes and part presence.


This is a photograph of a card sorting demonstration where a shuffled deck of cards was sorted by suit using machine vision. I believe the camera was a solid state GE TN2000 and GE Syracuse provided the vision controls for this one. Nearly all television cameras at the time incorporated vidicon tubes to collect the image data focused on it. When solid state cameras started to come out, the were all labeled CCDs, or charge coupled devices. When light was focused on the individual picture elements of the imaging chip, a charge would build up and would be read by scanning the array of elements and convert the charge to a voltage. By reading the array of charges, a video signal was produced. One problem with CCDs was that very bright light would cause overcharging of the picture elements causing them to bleed over to surrounding elements resulting in blooming, or washing out of large areas of the image. The GE cameras worked on a different principal and was named a charge injection device (CID). With the CID, all the elements were electrically pre-charged and light would cause the elements to discharge. Individual elements were not capable or drawing the charge from adjacent elements and therefore there was no blooming.


This is a demonstration robot we liked to call AP-C2, (Auto-Place See Too). (Star Wars came out in 1977) I think this picture was taken in late 1977 to early 1978. It was first demonstrated being controlled by an Imsai computer with a Z80 processor and 16K of Ram. Programming was done in assembly language for all the drivers and Zapple Basic for the main program. Ray Moran designed all the electronics and did the assembly programming, I did most of the basic programming, mechanics, etc. I'll bet Ray remembers what E900 means.


This is the prototype board for testing the electronics for the vision system that was the beginning of Opto-Sense. I have a tremendous amount of information and photos on all this, but I need to keep reminding myself that this is a web page, not a book.


One of the first customers for Opto-Sense (after AMC Jeep) was General Motors Fisher Body. Ed Woycik was their engineer who worked with us on several applications. He helped with the packaging if the electronics to meet the automotive requirements. The above unit was destined for the Grand Blanc, Michigan plant. Many of them were deployed to detect the presence of "clinch nuts" that were stamped into the inner sheet metal stampings on the trunk lids for the 'X' cars. When the clinch nuts were not present in the stamping, it would continue to be processed into a trunk lid assembly that couldn't be repaired. Opto-Sense detected defective inner panels in process and in time to be repaired saving money in scrap costs. Many of these units paid for themselves in 4 weeks based on scrap savings.


The above left photo shows a partial view of the inside of the cabinet. Each Opto-Sense unit of this type could be expanded to up to control up to 9 cameras from a single control cabinet. I designed the packaging details and laid out all the circuit board by hand using clear mylar with red, blue and black tape. Circuit board layout by hand is very tedious work. On the right is how the General Electric TN2500 Camera was packaged on an aluminum plate with cast aluminum cover shown behind.

Back To Home