But the laser range scanner they use costs > 5000 $ (!!!!)
formatting link
So, found some paper on the web that uses one webcam and laser pointer, but do not like pointing lasers around, from a flying plane would be dangerous.
The guys at MIT use an Atom processor, so with all that power why not use TWO webcams:
A quick web search showed me that I was not the only one with that idea:
formatting link
(Mrovlje).pdf&ei=WHkrUPeKPPC10QWc6IGICQ&usg=AFQjCNGXv_xwzCz1I-eBYxYB23emqDVD_Q&cad=rja (Distance measuring based on stereoscopic pictures)
This has restrictions, as he assumes cameras are perfectly horizontal, but whatever. I wanted to use the mpeg motion vectors, anyways I taped together a test setup and wrote some soft, and YES
Its actually EASY. And no reason why you cannot do an distance estimate of each 16x16 block in a picture at 10fps. Of course these distances are only valid in the plane through the camera lenses.
This is the eyes:
formatting link
These are my test pictures
formatting link
formatting link
those !%$!$#&!&@# Logitech webcams suck in bright daylight, and the fact that video4linux sucks and did not work did not help either, so I am now used to write my own drivers for ANYTHING, but it helped a lot in this test as only interesting part has any other data than grey:-) As you can see I did not bother to align the cams...
I asked it to find a match for the computah monitor in the picture, using 16x16 blocks,
******** found match at y=19 x=27 ********
So now we can calculate the distance from the distance between the monitors in the pictures.
I will have to work on this code, just a proof of concept.
5000 dollar? I think not :-)
So, but I am not sure anybody in sci.psygo^H^H^H^Hphysics still knows math.
I wonder if correlation functions per pixel are more accurate than spatial Fourier transforms. Each has its own susceptibility to distortion, paralax, etc.
The mind boggles at all the potential 'solustion' techniques.
On a sunny day (Wed, 15 Aug 2012 07:00:19 -0700 (PDT)) it happened Robert Macy wrote in :
Oh, you just opened a can of worms. I did, in the very long ago (10 years count?) some playing with the fftw library and Fourier for video resizing:
formatting link
Indeed you could make a frequency spectrum and sort of look for similarities. Had not thought about that. Now I just subtract matrices and take the absolute values of the difference, add all up for a 16x16 block, and store in a new matrix. Then look, after scanning, for the best match [1].
I did draw a little rectangle on the screen and put a 1 second delay in it, so I could see it working, while printing out the data on the terminal. makes it so much easier to develop and debug these things.
You could move one camera horizontally left and right from center, and look for a perfect line match. Moving mirror method would perhaps be possible. Polygon
Indeed there are X solutions. My prescalers arrived, but with 30 C to hot for soldering, so that is why I decided to play some software I guess.
[1] This means you could print a distance number on each square of the picture. Single digit fits just in it, meters perhaps.
I need all the GHz stuff for my radio altimeter (bit of a copycat design), but I really think this would work just as well, as the lower you go the better the resolution, just what you need, unlike a time of flight system that always has the same resolution Ideas ideas ideas.
On a sunny day (Thu, 16 Aug 2012 12:44:58 -0700) it happened Robert Baer wrote in :
I guess it all depends what you want to use it for.
I have been trying some modifications of the software, and some different picture material (some made with gimp) to test, and you can make pictures to cheat any system I think. It occurred to me that if you have ONE web cam looking down from a plane with known speed, you can calculate height from the frames in the movie (say at 30 fps jpeg), but even that can be fooled, say you fly over sea, easy for the algo to get confused on surface wave patterns I think.
It is one of those little fun ideas that once you get deeper in it, it could easily become a multi year project :-) In that case the 5000$ for a laser scanner could be a better choice. I am going to try some more stuff, what I have now is already very fast.
I think this is in the$45k range or so, but still an intereting solution used by military and artificial intelligence ?? for 'vision'
The original compnay made the best bass speaker systems ever, started in Sunnyvale, where I met the owner/founder John Hall and he completely described how he made the speaker! He knows his stuff! Solving all the technical details played like a chess game -- this does that, so we do such and such, and that causes this, so we do thingy, and so on. I thought my head would explode. I don't know how the company made the transition to a major lidar producer, but it did.Perhapd, he was looking for visual positional feedback of the whole cone. But, as he had explained, position of the cone does not always determine launching energy into the air.
Michael Dunbar - Business Development Manager Lidar Products
408-465-2859 "Mike Dunbar" mdunbar .AT. velodyne .DOTj. com
Velodyne 408-465-2800
345 Digital Drive fax: 408-779-9227 Morgan Hill, CA 95037
Lidar Products
Principles of Operation The HDL-64E operates on a rather simple premise: instead of a single laser firing through a rotating mirror, 64 lasers are mounted on upper and lower blocks of 32 lasers each and the entire unit spins. This design allows for 64 separate lasers to each fire thousands of times per second, providing far more data points per second and a much richer point cloud than conventional designs. The 64 lasers are employed with each laser/detector pair precisely aligned at predetermined vertical angles, resulting in a 26.8 degree vertical FOV. By spinning the entire unit at speeds up to 900RPM (15 Hz), a 360 degree FOV is inherently delivered. Regardless of the spin rate, 1.5 million data points (i.e. pixels) are generated each second, providing an exponentially richer point cloud than ever before possible.
The HDL-64E supplies returns out to 120 meters. It features ~1 inch distance accuracy and excellent repeatability. Radial resolution is dictated by spin rate, with radial accuracy as precise as .05 degrees. Additionally, state-of-the-art signal processing and waveform analysis are employed to provide high accuracy, extended distance sensing and intensity data.
HDL-64E Output Data Details The HDL-64E outputs UDP Ethernet packets at 100 Mbps. Interpretation of the data is fully explained in the user's manual, and each unit comes with a customized calibration file that guarantees ultra-high precision on both range and angular resolution.
Michael Dunbar - Business Development Manager Lidar Products
408-465-2859 "Mike Dunbar" mdunbar .AT. velodyne .DOTj. com
Velodyne 408-465-2800
345 Digital Drive fax: 408-779-9227 Morgan Hill, CA 95037
ElectronDepot 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.