Edge detection using LED light source.

[Nexvotum]

Here's my current project goal ( And what I need help with):

I want to use an LED light source, shine it through a 1" x 2" slit and then use a linear array camera to detect the edges of the slit.

The problem: Since LEDs are a point source of light emission, my camera has a difficult time detecting the edges of the slit. I use plastic as a diffusion gradient but it only seems to work so much.

Any ideas?

 

[NascentOxygen]

Hello Nexvotum! http://img96.imageshack.us/img96/5725/red5e5etimes5e5e45e5e25.gif

Those solar-powered LED garden lights use a clever diffuser to spread their light 360°. If the loss of intensity is not a problem, spend $2 and pull one of those apart to see whether you would be able to make use of that.

 

[Mike_In_Plano]

A single round collimation lens?

 

[sophiecentaur]

It is not easy to produce a uniform light field. One method I used, years ago, to give a really good illumination on a 35mm slide was to use a 'light tunnel'.

Imagine a long (say 200mm) square sided tube (the size of field you want) made up of four mirrors. You put your LED array at one end and a diffuser the other end. Light takes many different paths from each LED to the diffuser (think of a kaleidoscope) and you get a very uniform illumination across the diffuser. (You get two diffusion effects in tandem).
Construction is very non-critical (cardboard and tape). You can cut a thin, cheap, mirror or even use plastic mirror film in the internal faces.
My application was for an analogue video 'key' to inlay one video picture into another, using the video level on a third (a 35mm slide), to provide the switching waveform. The problem was to get rid of the 'contouring' on the 'control slide' which a simple condenser system with a 'coiled' filament bulb kept producing. This system gave a very convincing illumination across the area of the slide. It also gives a good light level, compared with a system which relies of a reduced aperture.

 

[Mike_In_Plano]

Cool technique. It seems we get into these difficulties with back lighting all the time. I've seen fibers that are woven and come together to have an LED on one end. Also, there's the usage of multiple LEDs behind a grating or electroluminescent material, though I find ELs have bright and dark places.

 

[NascentOxygen]

It's not clear whether devising the lighting is part of OP's project, but if not then maybe place the the object in front of a computer monitor (or television) displaying a white screen.

 

[sophiecentaur]

Cool technique. It seems we get into these difficulties with back lighting all the time. I've seen fibers that are woven and come together to have an LED on one end. Also, there's the usage of multiple LEDs behind a grating or electroluminescent material, though I find ELs have bright and dark places.

It all depends on how uniform you want your lighting. If you are edge detecting then it will not matter if there are broad, diffuse lighter or darker patches.

 

[hassou7]

If you are willing to detect edges, then you should collimate light to avoid perspective errors.
Don't forget that your LED is not a point source, but a very small array of them. Now, to make it a point source, either you use a multimode fiber, or you use a telecentric illumination by putting an aperture in front of your LED and in the ffl of your lens.
It is another way to do things if you don't worry much about the irradiance loss or phase.

 

[sophiecentaur]

It all depends on how uniform you want your lighting. If you are edge detecting then it will not matter if there are broad, diffuse lighter or darker patches.

That's a good point. You don't care about low spatial frequencies for edges.