It looks like you are using a couple of APC propellers. How it your machine performing. I have measured some propeller performance data for APC props for electrics up to 16" I also have some software to match props, gearboxes, motors and battery packs.
I found the name of the paper on the coaxial propellers, unfortunately, I did not keep a copy of the document itself.
Here are the details:
'An Experimental Investigation of Low Speed Single and Dual Rotating Propellers', Ronald Helmut Lorenz, University of Toronto, MSc Thesis, Department of Aerospace Science and Engineering.
I believe that it can be obtained through the library at the following URL: http://cisti-icist.nrc-cnrc.gc.ca/main_e.html
If you would like me to prepare a report for any specific combinations of propeller, motor, cells, please go to my website www.flightsolver.com and select the link for 'database summary' which is grouped with two other links near the top of the page.
I believe your comment that the second propeller should have a higher pitch is correct. The difficulty is determining how much. Another question arises, ie, is it better to have a space between two coaxial propellers.
I searched through the ~1,900 presolved powertrains generated by the FlightSolver program. These powertrains are generated by the FlightSolver program according to minimum performance criteria and battery and motor current and voltage limits for use with airplanes.
The APC 12x8P and APC 12 x 6E are not in the database, but I have the APC 12 x 8E (tractor) propeller in the database.
I found a presolved powertrain with a APC 12x8E that reached about 4.7 grams/Watt at zero airspeed (ie in a hover). The thrust in a hover is 17 N and the powertrain mass is 0.37 kg (3.7N), so the powetrain can lift its own weight.
The current drawn by the motor in a hover is about 31 Amp which slightly exceeds the battery pack current limit of 30 Amp. Flight time is thus quite low, about 2 minutes.
The complete powertrain is as follows:
propeller: APC 12 x 8E
motor: AXI 2826/12
battery: FlightPower (UK) EVO20 1500 4S1P
I have the complete report generated by FlightSolver in pdf form. I can't figure out how to attach it to this email, however, I will figure it out and send it along in a separate email.
did you make any progress with your sparse stereo? I was checking your page for some sample images.
I'm currently trying to implement something pretty similar. Eventually the algorithm is supposed to track features in consecutive frames from one camera but as it doesn't really come at additional efforts i'm currently using it to match features between the frames from a stereo pair.
My current approach is a fixed point implementation of the harris corner detector in connection with a sum of absolute difference matching of some neighborhood around detected features (though normalized cross correlation might be equally fast and more reliable). Though i'm currently using non rectified images (it takes forever to load the distortion tables onto the board) the first results look promising.
I would be glad to hear about your progress and ideas.
Yes, we are getting reasonable results with the sparse stereo search. Video from a very early version of firmware was captured here- http://www.surveyor.com/cgi-bin/robot_journal.cgi/2009/07/26#224 and the firmware has improved quite a bit since. We are using a giant lookup table for rectification that is generated from polynomials, so there isn't much storage required.
those results look quite promising! What feature detector do you use? It looks like it is a little more sensitive to edges than corners. (Which shouldn't be a problem for calibrated stereo rigs but might be one for monoSLAM applications.)
I really like the idea of generating the lookup table on the processor rather than storing it all in memory. Currently i'm generating that table in Matlab.
Is there any video showing the matching too? I would be really interested in that.