I need to project a number of near-infrared spots on a ceiling that a robot can see, and I need to keep the total parts cost low while keeping the spot as bright as I can manage. Size of the spot is not critical; my detector (photodiode)n sees a one inch spot and a one foot spot as being the same if the total illumination is the same.
The obvious solution is to pick whatever infrared LED gives me the most light for the buck, and to use multiple LEDs to get whatever brightness my detector needs. Can anyone here think of any other ways? The only alternative that I have been able to come up with is a laser diode.
This will depend strongly on the size of spot you wish to project. If you can get some cheap plastic lenses, you can greatly focus the beam from an IR LED into a very respectable dot. However, with a single lens the best you will get will be the image of the LED die itself. If you have a friend who repairs CD-ROM drives, you can salvage some very nice IR lasers. These can also prove hazardous, though, so take care with eyes and IR lasers! Another trick I have tried (and it works fairly well) is to put a small bundle of large plastic light fibers in front of the IR LED and then break them out into individual little "dots". These can produce far smaller light spots and provide numerous sources from a single LED.
I was just investigating this very subject. However, in my case I wanted to project the dots onto a 3-dimensional object so that I can take pictures of the dots and use with a photogrammetry program.
I was planning on using a near-infrared laser pointer (Laser diode modules are available in true-infrared if you prefere). The wattage was well under 5mw and so safety would not be a concern in my situation.
Of course, the laser makes one dot. But there are optical devices called diffraction gratings. These will divide that dot into several dots. I was going to use two of them... one to make the horizontal row of dots, and the other at 90degrees to make the vertical columns.
Here is a link to one site that sells some holographic-type pattern generators for only $10
With laser pointers so cheap these days (under $5 each, and less if you know where to look) a practical and inexpensive approach is just to use multile lasers. You can PWM pulse them to keep the current demands low, though they don't consume much power anyway.
A triad of pointer lasers produce three very bright spots, much brighter than the main/sub beams produced by a diffraction gating. I found that the cheaper cameras could not reliably "see" the second- and sub-order beams, but could readily see the triad. Scanning the triad at least in the X axis, like with a small servo, can scan a fairly large scene, and it's easy to sync the start of each line. A small DC motor fed with a sinewave should work fairly well as a galvo (with appropriate damping to limit rotation), though I haven't tried this yet.
I first wrote about using a diffraction grating and laser to determine depth and contour in a book I did in the mid-80s. I have not heard of anyone who replicated the experiment for vision. If you do, please share your results!
As are flying spot and other methods. But the idea of a diffraction grating is that the number of sub-beams is controllable by the fineness of the grating, and that topology of an entire 3D scene (rather than a line at a time) can be read by measuring distance between dots -- assuming the use of a grating that is scribed in both the X and Y planes. It's a real-time (e.g. whole frame) method of vision analysis. You do need a fairly strong laser and a good camera, though.
Even though some guy patented the idea in the late 80s, and there are some commercial products based on it (even CD focusing works this way) I have yet to see the concept used in an amateur robotics setting. Or maybe it has -- just hven't looked in the right place.
I've been waiting to get the right imager for this. I have the laser and diffraction grating, because of _The Robot Builder's Bonanza_.
With just the center and two sets of harmonics you can determine distance, contor, and a variety of other things. It gets easier if you use three difraction gratings so that you get a hexagonal pattern rather than just a square pattern.
It should be possible to make a very tidy representation of the surface with only a few numbers. These could be fed into both a neural network and a behavioral algorithm for various types of responses.
Since the difraction of the laer happens in a conic section, the distance can be calculated by noting the distance between the points. The further apart the points, the further the obsticle is. The brightness of the points can tell you something about the color of the texture.
The angle of the obsticle can be seen by the distortion of the pattern.
If one or more of the points are not where calculated, then the obsticle is uneven. Exactly where the point can be determined.
It is best if you know the exact angle and distance between the laser and the imager.
If I understand your question, perhaps at longer projected distances this is true, but this type of vision system would be for close proximity -- under 10 feet maybe. Assuming the selection of the proper grating (not too fine, not too coarse), the beam separation will be greater than the vanishing perspective. That is, one doesn't have to cancel out the other.
Do note that every CD player now made works the same way. The beam of a laser is split by a diffraction grating that is scribed in one plane. This produces a single bright spot in the center, and multiple "sub beams" to the side. In a CD player, only the two spots to the immediate right/left of the main spot is critical; the rest are ignored.
The reflected beam is then directed to a set of photodiodes. The variable distance between the sub beams affects the amount of light falling on each diode. The player adjusts its lens to make the beams more-or-less equal on the diodes, signifying proper focus. So, in this microscopive view situation, the net effect of the beam separation is greater than any changes in the reflected beam pattern as the disc moves closer or further away from the lens.
This sounds like a really interesting approach as I've been trying to implement a simple (ya right!) laser ranger finder with a PSD device but it's a pain-in-the-neck to build as it requires precise alignment of the optics and I'm no expert!
Where would one get some of those gratings? Or is it possible to build one easily. Keep in mind that I don't need anything high precision here.
Being able to project two divergent dots with a single laser and calculating the distance using a simple camera would be real simple if I could simulate the grating effect. unless it can be done with somethign else? Using two lasers???
There's no reason you need a diffraction grating. As Gordon said, you might be better off with multiple lasers. They'll be brighter and they only cost a few dollars each.
The diffraction grating I used came from a pair of 50 cent "rainbow glasses" I bought at a science museum gift shop. These are scribed in both X and Y, which gives a series of both horizontal and vertical dots. If all you need are dots along one plane, then you can get a suitable grating from a junked CD player. You can also make a laser diffractor using a knife edge, which can be as simple as an X-acto blade glued to a clear plastic aperture placed in front of the laser.
When I played with this I mostly used 10 mW he-ne lasers. They are brighter, and the beam waist is larger, than common diode lasers. If the dots are fairly weak, then your camera (filtered or not) may not be able to see them against the general lighting of the scene.
Thanks for the great ideas Gord. Call me cheap but I always try to find the home-brew solution before buying expensive specialized equipment/tools. Take for example a laser beam generator optic: I'm sure it's not that expensive and all but played with different pieces of plastic and glass until I realised that a glass fuse does the job wonderfully. Then I asked a friend of mine to get me one of those glass stir stick they use in a lab and it works even better! That's my optic tip of the day! :)
Now if I could only find a way to "make" a devent laser filter. Any ideas?
PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.