az110597
asked on
machine vision system design
I need to design a machine vision system to inspect molded plastic
tail light covers. These covers are made from red translucent plastic. They are
approximately 70 x120 mm in size. Like most tail light covers, these have
a regular arrangment of small pyramid structure on their back sides. These
small structure allow the covers to reflect light so that head lights
hitting the covers will cause a reflection even when the tail and/or brake
light is not on. All the covers to be examined are of one design. You are
to find discolorations and cracks. You must also make sure that the four
holes that allow the tail light to attached to the car are all present at
the appropriate location and are of the appropriate dimension. Finally
you should be able to detect areas that might not have been molded
correctly and are too thin. While the rotational orientation of the covers
cannot be controlled, they will always have the same surface facing up.
What hardware, imaging a nd lighting geometry and algorithms do I need?
tail light covers. These covers are made from red translucent plastic. They are
approximately 70 x120 mm in size. Like most tail light covers, these have
a regular arrangment of small pyramid structure on their back sides. These
small structure allow the covers to reflect light so that head lights
hitting the covers will cause a reflection even when the tail and/or brake
light is not on. All the covers to be examined are of one design. You are
to find discolorations and cracks. You must also make sure that the four
holes that allow the tail light to attached to the car are all present at
the appropriate location and are of the appropriate dimension. Finally
you should be able to detect areas that might not have been molded
correctly and are too thin. While the rotational orientation of the covers
cannot be controlled, they will always have the same surface facing up.
What hardware, imaging a nd lighting geometry and algorithms do I need?
thats a 5.000 US$-question, not a 200 points one!
You can take the mars lander approach. You don't need to inspect for all the details, you only need a backlight reference measurement off a good lens as a comparison. Then, position each lens in front of a standard light source, backlight each lens, read its composit image with a photocell array, and compare it to the reference good lens image. Calculate the degree of difference between reference good image and current snapshot. Use multiple test runs to set the sensitivity, then reject any lenses outside of the acceptable limit. This will work for your install holes as well since the light shining through the holes will vary with position and hole size.
Your sensor and readouts are your main problem. Try the cheap things first, like a personal scanning system!
Your sensor and readouts are your main problem. Try the cheap things first, like a personal scanning system!
ASKER
I need to be able to evaluate the type of failure: whether it was
due to discoloration, crack, uneven thickness, improper holes,
misaligned pyramid structure, etc. I also need to know the various types of algorithms.
due to discoloration, crack, uneven thickness, improper holes,
misaligned pyramid structure, etc. I also need to know the various types of algorithms.
ASKER
This is a 200 point question so the answer should be long
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.