<< return to Pixycam.com

Pattern recognition with Pixy

Hi,
RoboCup Junior Recue Maze 2017 has a new challenge addressing the age group 14-18 years. The challenge is to build a robot which traverses a maze and recognizes “visual victims”. Each such “victim” is a single letter, attached to the wall, printed in black, Arial typeface, height 4cm, and it can be only one of the three letters “H”, “S” and “U” (http://rcj.robocup.org/rcj2017/rescue_maze_2017.pdf).
I wonder whether it would be possible to use CMUcam5 Pixy for that, and what would be the necessary steps (i.e., how to write own software for such a task).
Any help greatly appreciated,
Holger

Hi Holger,
Thanks for the message. Pixy doesn’t do this kind of recognition currently, but Pixy could do so with some modified firmware:

http://cmucam.org/projects/cmucam5/wiki/Building_Pixy_Firmware

Hope this helps!

Edward

Dear Edward,
Thank you very much for the link; however, for beginners it is not very helpful.
Is there any “reference project” where somebody has implemented custom-made
pixy firmware with the Keil MDK-ARM SDK or LPCXpresso?
What would be the appropriate steps to start with?
Are there any image processing libraries which we could use?
Regards,
Holger

Hello Holger,
For people just getting started in computer vision, developing custom algorithms, etc, there’s perhaps OpenCV, but it’s not beginner-friendly per se, but it’s easier than developing computer vision algorithms on an embedded system (like Pixy). Unfortunately, computer vision custom algorithm development is a difficult topic and beginner friendly material is hard to find. Sorry I couldn’t be more help.

Edward