search

IEEERobotics / bot

Robot code for 2014.
BSD 2-Clause "Simplified" License
17 stars 10 forks source link

Using OpenCV for localization. #442

Open AhmedSamara opened 8 years ago

AhmedSamara commented 8 years ago

After much struggling, it seems that we won't be able to use the modified QRCodeStateEstimation library on the final robot, due to the ZBAR library being incompatible with the BeagleBone, as well as our struggling to get BOOST library working to work.

Looking at the source code for that library though, it's not that complicated and we should be easily recreated in python.
All you really need is a set of (known) points, and to be able to identify them in the images captured by the webcam.

This is the part that made Zbar really simple. When it finds a QR code, it also tells you where the vertices are, and this is the part that I'm running into problems with now.

I was hoping to solve this problem using SURF points

The change in how we're approaching this is explained in my senior design presentation:

opencv

Basically the problem I'm running into is that even though I can now identify a bunch of known points on the QR code, there's still a few problems:

AhmedSamara commented 8 years ago

Found this:

http://docs.opencv.org/3.0-beta/doc/py_tutorials/py_feature2d/py_matcher/py_matcher.html

AhmedSamara commented 8 years ago

This looks promising now

https://gist.github.com/AhmedSamara/0ce6a836dd45e16b68ba

AhmedSamara commented 8 years ago

matches

It looks pretty promising.

PaladinEng commented 8 years ago

Is this running on the BBB?

On Sun, Jan 24, 2016 at 5:43 PM, Ahmed Samara notifications@github.com wrote:

[image: matches] https://cloud.githubusercontent.com/assets/2405319/12539494/ebbd985c-c2c1-11e5-92b3-af8da39d49a5.png

It looks pretty promising.

— Reply to this email directly or view it on GitHub https://github.com/IEEERobotics/bot/issues/442#issuecomment-174349006.

AhmedSamara commented 8 years ago

http://stackoverflow.com/questions/7263621/how-to-find-corners-on-a-image-using-opencv

Here's another strategy for generic objects.

AhmedSamara commented 8 years ago

Two cameras might help, here are instructions for setting up the second:

http://ubuntuforums.org/showthread.php?t=168221