<< return to Pixycam.com

Pixy Linux V4L Device

Pixy Software
#1

Hello, and good afternoon!

I want to know some information in how does Pixy work with Linux. I want to stream it to ROS as a V4L device, is this do-able? How does the PixyMon software manage this on Linux? It’s a proprietary raw image gathering from USB?

Thank you in advance!

Pedro

#2

Hello Pedro,
Pixy doesn’t use V4L in Linux. It uses a simple custom protocol ontop of libusb. Libpixyusb is a straightforward library for interacting with Pixy and grabbing images over USB:

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

Here’s some example code on how to grab a raw image using libpixyusb:

http://cmucam.org/boards/9/topics/5106?r=6183#message-6183

Hope this helps!

Edward

#3

Hello Edward!

It helped! But now I am asking myself if there is any interface for C++ instead of C, or even Python. I guess not, right?

Pedro

Edit: Also, I would like to know what resolution I can take from the camera interface, through USB. As I’ve seen, most of the detection gives back coordinates in a 320x200 space, and I would like to have them in a 640x400 space. Is this do-able?

#4

Hello Pedro,
libpixyusb is compatible with C++. Here’s some info on Python:

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

Regarding resolution, grabbing raw camera frames is limited to 320x200 because Pixy doesn’t have enough free RAM to store a larger frame. Sorry… :frowning:

Can you describe your application? Pixy can be used as a frame-grabber, but Pixy works best when the image processing is handled onboard Pixy.

Edward

#5

My application is to use a Pixy to get positions from 4 IR LED’S which are 30cm apart from one-another and mounted on a UAV which can be at up to 5m/4m from the camera. Do you think it will work on these conditions?

Thank you in advance!

Pedro

#6

Hi Pedro,
Your application sounds very similar to what the IR-lock folks are doing

You might check out what they’re doing and how they are getting positions of IR sources.

Edward

#7

I have asked them if it can detect 4 LED’s 30cm apart from each-other and at 5m (3m would be do-able) distance from the sensor, but the main problem is if it can distinguish them nicely. Perhaps you could give me some hint on this?

Thank you in advance Edward!

Pedro

#8

Hello Pedro,
By distinguish them, do you mean detect them? Or do you mean telling LED 1 from LED 2? Detection is solved very robustly by the IR-lock folks. Telling the difference between LED 1 and LED 2 is a more complicated problem though. I’m not sure what the best approach would be for solving this.

Edward

#9

I mean detect them. I need, on the output, something like “there is one LED at x and y, there is another LED at x2,y2, …”, but the thing is, with low resolution, at a certain distance (5m is my objective), does it detect the 4 LED’s considering they are 30cm apart from each-other in the body frame?

#10

Hello Pedro,

does it detect the 4 LED’s considering they are 30cm apart from each-other in the body frame?

At 5M, it seems reasonable. You can calculate, assuming a 70 degree horizontal field-of-view.

2*tan(FOV/2)*d = Width

2*tan(70/2)*5m = 7m

7m/640 = 11cm per pixel

If you go with the IR-lock lens and algorithm, the field-of-view will be different. It will probably be narrower. For your application, you might consider using a longer focal length lens since your objects are so far away.

Edward