suggestions on non-contact distance sensor

ktekx

I'm currently in the planning/brainstorming stages of my senior project and one of my idea requires very precise measurements of an object distance. I guess what I am looking for is some sort of "non-contact micrometer" so to speak. It could be within +/- several millimeters resolution accuracy, it doesn't have to be THAT precise since this is a proof of concept, however, we are limited to a budget and all of the laser types are wayyy out of my price range.

I was wondering if anybody knows of an alternative that is fairly accurate? What about those LEDs used as sensors, I saw this video a while back and thought this could work: www.youtube.com/watch?v=7kzIHMpOt20&feature=related

Any help would be appreciated, thanks.

LoopyByteloose

Each technology has its own shortcomings.
LEDs are limited by the color of the object these measure. In some cases, they just don't see things. One might have to use an array of colors to see all objects in one device.

Ultrasonics are very go and can go much further distances, but in the atmosphere distances are effected by humidity and wind effecting the speed of sound. That may not be an issue at micro distances and one can certain enhance calculations by reading the humidity as well.

Thermopiles and PIR need a contrast in temperature and possibly movement as a que.

Ferromagnetic objects can be recognized quite easily and metal detection was some of the eariest technology to develop.

I suspect with laser pointers being so cheap, laser are quite affordable. Recently there was a neat little schematic of using the laser diode with a 2n2222 transistor controlling is intensity from a regulated 5 volt supply. That might be all you need on the output side. By using appropriate photodiodes [noparse][[/noparse]which are much faster photo sensors], you could do a lot.

So you need to set up parameters of distance range, object type, and degree of accuracy in order to choose which. Otherwise you may waste a lot of time with the wrong technology.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
It's sunny and warm here. It is always sunny and warm here.... (unless a typhoon blows through).

Tropically, G. Herzog [noparse][[/noparse] 黃鶴 ] in Taiwan

Phil Pilgrim (PhiPi)

kyekx,

You didn't mention what the maximum range has to be.

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Still some PropSTICK Kit bare PCBs left!

ElectricAye

What sort of surface material are you trying to detect? It probably makes a big difference. Also, what is the surface finish? Is it smooth, bumpy, fluid, fibrous, etc?

ktekx

Sorry, I think I should have been more clear on what my requirements are. I need to measure the surface of common objects to gauge the depth of the surfaces to recreate it in 3d. Sort of how 3d artists can replicate a real world object using a feeler gauge. I would like it to be contact-less, though. Is it at all feasible with LEDs (such as the video) if the color/surface of the object is kept static, ambient light not being a factor, etc?

Thanks for the quick replies, guys

ElectricAye

The color of the light and the transparency of the objects will affect your readings. For example, flesh reflects green light much better than red. Red light tends to penetrate deeper into the skin, which is why black and white photographs taken through green filters will make people look kinda like they've been carved out of stone: the green light coming off their face has been reflected only from the uppermost layer of the skin, which is a mostly dead layer. In contrast, oil paintings have a rich glow about them because the light penetrates and interacts on several layers, giving the images a somewhat fleshly, living glow. That also explains why wax is used in wax museums. So if you're looking at sub-millimeter accuracy, your measurements will be affected by how deep the particular frequency of light is going into the material. Generally speaking, the shorter (more blue) the light, the less deep it will penetrate. Dark, frizzy materials would give you the biggest problem using techniques similar to what you see in the aforementioned video. I'm guessing what they use is some kind of an array of intensity sensors matrixed between the LEDs. In that case, another type of object that could give you fits is one that is highly reflective: if the light emitted from a particular LED bounces off at a 45 degree angle, the intensity sensor might not ever see it again, and instead an adjacent sensor would pick up that signal and get confused.

So, I'm just guessing here, step one to improving what you see in the video is to go with green or blue LEDs.

Post Edited (ElectricAye) : 9/11/2008 4:52:38 PM GMT

ktekx

Kramer:
I was thinking that if the led is calibrated to the object, it may be able to read a fairly accurate measurement?
PIR and sonic sensors maybe not be accurate enough, I think, although I have not used ultrasonic sensors though, what are their resolutions like?

The laser micrometer I was speaking of was something like this
www.micro-epsilon.com/staticcontent/images/sensoren/optoCONTROL_2500_-_Laser-Mikrometer.jpg
They are too expensive (way out of the allowed project budget) and I cannot use them.


ElectricAye:
I was thinking that for the proof of concept of my design, to use a preselected object rather than anything so that it would work most effectively for the sensors. I'm just wondering how high of a resolution I can obtain from leds and also how narrow of "pin-point" I can achieve, since I want to detect points on an object to render them back into 3d.

ElectricAye

ktekx,
Answering this question is probably harder than you thought. From the image you showed of the laser system, I would say your main problem is how to collimate the light coming from an LED. Your everyday LED spreads out light quite a bit. And collimating light (getting it to travel in a nice straight line) is not as easy as it might seem, especially when you must do it in a tiny space. For example, look at the sketch I've attached. Simply putting a pinhole in front of the LED does not generate a "cylinder" of light the same diameter as the pinhole. Notice the dotted lines and how the path of the light from the extreme regions of the LED ends up crisscrossing in the pinhole and spreading out by the time you get to the object and the observer. Sticking the pinhole right up against the LED does not solve the problem. What can help is to place the pinhole far away from the LED light source, but then your sensor gets longer and longer and your light source appears to get dimmer, too. If you make the pinhole extremely small, then diffraction effects can start to make things messy and make things dim as well.

I'm not an expert on this sort of thing, but I'm guessing there might be a trick you could consider using. If you have a bunch of LEDs lined up on one side, and a bunch of sensors lined up on the other side, you might be able to turn on one LED and then scan through the entire line of sensors and measure their outputs. Then turn off the first LED, turn on the second LED, and scan through the sensors again. By using this method of scanning through sensors and LEDs, perhaps you could use software to "see" the edge positions more accurately than you could with a "straight shot" approach. I think the technique is called deconvolution. It's a method of getting sharper data out of instruments that are sorta blurry-eyed.

I hope that helps,
Mark

Beau Schwabe

ktekx,

A classic Wheatstone bridge might work for non-ferrous as well as ferrous objects.

en.wikipedia.org/wiki/Wheatstone_bridge

I have attached a capacitive version. The method of operation is basically the same.

A differential sensor approach (using two legs of the Wheatstone bridge) will give you a relative angle that an object is to the surface of the sensor and negate "lift-off" effects.
On the other hand a single sensor approach (using one leg of the Wheatstone bridge) will give you relative distance. To be an effective sensor for this application you probably
would want both differential and single mode capabilities.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.

ElectricAye

I suppose another approach would be to use a sensor mounted on some kind of moving arm. The arm's only mission in life is to get as close to possible to the object without touching it. Your Prop keeps track of where the tip of the sensor is via counting pulses to a stepper motor, from an encoder, etc. See the following link for discussions and sketch...

http://forums.parallax.com/forums/default.aspx?f=25&m=285531&g=285987#m285987

LoopyByteloose

PIRs are mainly motion detectors. Your project is far more precise. Originally I thought you were concerned with collision avoidiance.

Murata has a large variety of piezoelectric ultrasonic devices. Considering that ultrasonic is used in medical imaging of internal organs, it may or may not be useful for your particular context. Some of Murata's devices are paired transmitter and receiver and look like the devices that Parallax's Ping unit uses. Still, Daventech has come out with an array of devices with different degrees of accuracy and different distance abilities.

I suspect the best of all would be to simply project a laser led on the object while mounted to a stepper motor driven turn table. The laser's height would also be controlled by a stepper motor. Then have not one, but two video camera's accumulate position data of the red dot that is correlated to the turntable and laser light source's height.

Of course, this is a lot of number crunching.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
It's sunny and warm here. It is always sunny and warm here.... (unless a typhoon blows through).

Tropically, G. Herzog [noparse][[/noparse] 黃鶴 ] in Taiwan