Optical Flow and New IMU board

Geof Barrows of Centeye sent me some raw sensors so I can develop for them. These are great sensors due to inpart there flexibility, low external part count (you can see the board has two capacitors), and the use of chip on board technology. They are fairly easy to use.

As of right now, I am able to read the image in log mode. Once I get a lens, then I can start some image processing code. First full image (8 bit greyscale):

I am streaming pixel data via a wireless modem as the USB connection on the XMOS is VERY slow, limiting the frame rate I can use it at. I am grabbing one frame in around 32 milliseconds, but I should be able to bring that down with tuning of delay values. The whole image only takes 32K RAM, meaning I can work with the whole image, and have 32K RAM for processing capabilities. Final plan consists of a scalable architecture with add on modules to support any number of sensors. The sensors will be trained via IMU.

In the world of IMU's, I ordered another board as the other was messed up beyond recovery (port mappings). I also messed up this time (Flipped two ports...), but it is usable.

I am using Bill's DCM algorithm, and I need to tune the PI values. Current output isn't very usable:

10,000 points:

You need to be a member of diydrones to add comments!

Join diydrones

Comments

GR0B August 23, 2010 at 6:02am

Ravi did you try the pinhole? I am interested on how well it worked.
Ravi Gaddipati August 18, 2010 at 3:28am

I will try a pinhole today!
GR0B August 17, 2010 at 6:57pm

You could make a basic Pinhole camera lens for testing with a little tin foil (useful for until you get some optics)
Geoffrey L. Barrows August 16, 2010 at 10:03am

Ravi, just for fun, try repeating the carbon fiber shadow experiment with just one change: set HSW to 254 and VSW to 127. This will form super pixels by shorting out 8x8 blocks of pixels. You can read out all 128x256 pixels like you are doing now, or just read out every eight row and column to form a 16x32 image.
Developer

Mark Colwell August 16, 2010 at 6:25am

Thanks for conductance/dielectric data for JB Weld, good to know. I always have some around !
Patrick Hammer August 16, 2010 at 2:50am

Is JB Weld conductive?

Nope – it has around 16% metal/iron powder in it, most of which is partly oxidized. We’ve measured its dielectric strength at work on several occasions – and it always comes in above 450 volts/mil (even after a day in a chamber @100% humidity) - it's plenty plenty safe on boards
Ravi Gaddipati August 15, 2010 at 4:27pm

I have working log readings:

This has a piece of carbon fiber right over the sensor (my program accommodates for the opposite response to light).
Im not sure if the vertical line is hardware or software. It being in the middle (col. 128) has me thinking its software.
Geoffrey L. Barrows August 14, 2010 at 7:11pm

I don't know about JB Weld but JB Kwik works fine for us. I haven't measured it's conductance, I guess it's possible it has a tiny conductance but we haven't noticed it yet. For gluing the chip to the PCB we use a silver conductive epoxy- more expensive!
Ravi Gaddipati August 14, 2010 at 7:08pm

Hey,
I will try that, after working with the linear mode.
I thought JB-Kwik and JB-Weld were conductive, but it seems there not! I will be using mine much more now :)
Update to come tomorrow if I can get linear to work.
Geoffrey L. Barrows August 14, 2010 at 6:54pm

Ravi- Try using an LED high above the chip to illuminate it, and then sweep a paper edge etc. close to the chip to make a moving shadow. You should get some OF then.

of 2

This reply was deleted.