<< return to Pixycam.com

White Goalposts detection for Robocup

Pixy Software
#1

Hello everyone! I’m new using Pixy but I think is a great powerful sensor. I have experience with C/C++ so programming should not be a problem.

I will participate in the Robocup(Humanoid). Pixy doesn’t have a problem detecting the ORANGE ball but this year there are new rules that the robots has to detect WHITE goalposts, last year were YELLOW. I’m a little lost here so I need your advice and help in how to recognize the white goalposts using Pixy.

  1. The Pixy will be connected to a Beaglebone so I dont know if I should change the firmware to detect white objects and lanes, or the Beaglebone should process the image.
  2. Does someone has a basic code for detecting white objects and lanes that could share (for firmware or for Beaglebone)?

Please give me your advice and any comment will be really appreciated.

Once the robot is finish I will share with you all a video how it was.

Thanks everyone in advance!

#2

Hello Abraham,
White is difficult because it’s so common. It shows up everywhere!

Do you have a link to the Robocup rules? maybe we can help!

Edward

#3

Thanks for the support Edward, the rules are in here

My plan is to detect white shapes, and look just for lines, with a dynamic thickness checker, and process them, I worked previously with a similar algorithm to make an autonomous follow lane car but I used OpenCV and LabView (easy). This time it should be integrated in the Pixy or in the Beaglebone. what do you think?

#4

Anyone else?

I need your advice and help in how to recognize the white goalposts using Pixy.

  1. The Pixy will be connected to a Beaglebone so I dont know if I should change the firmware to detect white objects and lanes, or the Beaglebone should process the image.
  2. Does someone has a basic code for detecting white objects and lanes that could share (for firmware or for Beaglebone)?

Please give me your advice and any comment will be really appreciated.

Once the robot is finish I will share with you all a video how it was.

Thanks everyone in advance!

#5

Hi Abraham, I have no experience with BB, but since you have already experience with OpenCV I would suggest to go with it or maybe use some computational software such Octave (if that is possible in BB). I would do definetely the processing with the board since it has more capacity for doing it. I did create a lane detection platform with the Hough transform but with Matlab (that is why I suggest you Octave). Regarding the white objects I have no idea, but if you have good knowledge on C++ I think it would not be a problem create a “white detection object”, but no clue how I would port it to Pixy (memory is limited).

And another question (and I am currently stuck here) is how to transfer the stream video coming from Pixy to the CPU that processes the video (in my case Matlab)

Hope this helps!

#6

You can save individual images to disk from PixyMon and then load these images into Matlab. This is what we’ve done in the past when we need to test algorithms in using tools like Matlab.

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

Edward