I was pleased to see a cool laser rangefinding project on Kickstarter- I hope this project gets fully funded (and I'm a backer). I've actually been experimenting myself with structured light and laser rangefinding using our ArduEye hardware and thought I'd share it here.
The setup is very simple- An Arduino Pro Mini serves as the computing backbone of the device. Via a 2N2222 transistor (I know I know...) the Arduino can on and off a red laser module. The Arduino is connected to an ArduEye breakout board with one of Centeye's Stonyman image sensor chips and a cell-phone camera lens. The whole setup (excluding the red FTDI thing) weighs about 10.9 grams. I think we can reduce that to maybe 4 or 5 grams- the laser module weighs 1.9 grams and is the limiting factor.
The principle of operation is straight forward- the laser is mounted horizontally from the image sensor by a known baseline distance. The Arduino first turns off the laser and then grabs a small image (3 rows of 32 pixels in this implementation). Then the Arduino turns the laser on and grabs the same pixels. The Arduino then determines which pixel experienced the greatest increase in light level due to the laser- that "winning point" is the detected location of the laser in the image. Using this location, the baseline distance, the lens focal length, the pitch between pixels on the image sensor, and basic trigonometry, we can then estimate the detected distance. I haven't yet implemented this final distance calculation- my main interest was seeing if the laser could be detected. The above video shows the system in operation.
In practice, I've been able to pick up the laser point at a distance of up to about 40 feet- not bad for a 2 mW laser. In brighter lights you can put an optical bandpass filter that lets through only laser light- with this the system works at distances of say 10 feet even in 1 klux environments e.g. a sunlit room. If you are using this for close ranges, you can turn up the pulse rate and grab distances at up to 200Hz. How does an Arduino grab and process images at 200Hz? Easy- at 3x32 it is only grabbing 96 pixels!
Comments
It's a joke about the 2N2222 (better known as the 2N-quad 2) as being one of the most common transistors on the market and in the professional electronics world if you use this type of "common part" then you are clearly an amateur. However, the 2N2222 is a superb part for many applications so there's nothing wrong with using it.
@Gary - fwiw the Sharp sensors are on the slow side, about 25Hz is what I think I calculated it to based on oscilloscope observation (I think the datasheet may actually have it specified).
I found another link to this same rangefinder and one of the responses on it indicated good results using a TAOS now AMS linear array sensor which should actually be a lot easier and faster to decode for the ranging function.
He was only using a 128 pixel linear array here http: //embeddedeye.com/profiles/blogs/11-gram-ardueye-driven-laser-range finder
And he was not using well developed optics and his results deteriorated badly with distance.
But AMS now has a 1280 pixel linear array for under $50.00 and with proper optical design that might provide a considerable improvement in range, resolution and update rate.
Just a question really, I am planning on using one of the longest range sharp IR ranging sensors on a servo scanned base to permit low resolution adaptive scanned depth image construction.
The Sharp ranging sensor uses intensity of IR dot reflection to produce a distance calculation.
The Sharp array can work out to 16 feet also and costs $26.00.
The distance calculation decreases in accuracy with distance, but this is true for the camera method as well.
And the update (frame) rate is considerably higher.
This is the Sharp IR sesor I am planning on using: http://www.robotshop.com/productinfo.aspx?pc=RB-Dem-05&lang=en-US
I did also notice that Parallax has a laser ranging camera system that works like yours but it appears to have an abysmally slow update rate of 3 or 4 times per second and it is huge.
I notice that the KickStarter project shows as fully funded by 3 times over that requested as of 6 months ago.
Any news from them?
@Monroe- I've never tinkered with LIDAR itself since it would involve accurate timing that I've never put into our chips. (I *could* do that, just *haven't* gone that direction yet.) But it is an interesting idea.
On the other hand, I've a few ideas for some low light sensors using SPADs (single photon detectors)...
@Marko- actually that is in the code and it works but I didn't report it here. I took the winning pixel, it's two neighbors, constructed a LaGrange polynomial with them, and computed the max. The only cost is that it works better if you blur the image slightly, which reduces the range and/or max light level the device can operate. Also the pixels have a log response so a different interpolation method might be better.
@Lefebvre- agreed. I think though with more tinkering I can prob get it working outside, certainly with a different image sensor or laser.
This is pretty neat, but IMO, severely limited if it can't work outside.
Wow, this looks quite promising.
I would think one could do filtering of the data to deal with any texture-induced noise... ?
200hz isn't quite enough to do a fast updating, high resolution XV-11 style LIDAR but it's way, way better than nothing. :)
-
1
-
2
of 2 Next