Hello FIRST teams,
I’m wondering if any of you have come across a nice USB camera for vision processing. This year, my team is planning to do vision processing in a Raspberry Pi, which has USB ports, which opens up nice low-cost alternatives.
Anyway, does anyone have link or model number to a sufficient camera?
Any decent webcam should work. You could also use the Kinect from last year’s KOP.
Fair enough. I’ll look for a webcam and post back results for anyone else.
My team doesn’t want to use the kinect because of weight concerns.
A Playstation 3 “Eye” camera produces good images and can run at 75 fps at VGA resolution.
You may want to review the Rasberry PI option … nice unit, great price but at 5 fps at 320x240 without any actually processing (i.e. this is just image capture) you may want to ensure that you’ll be moving VERY slowly.
Depending on what you are doing with it … try reducing to 160x120 which will help with the fps a bit. You should also ensure that the PI is running at top speed (there is a config file that sets the CPU speed)
You will not need the top of the line HD webcams. Anything around $20 should work just fine. Keep in mind that some drivers are easier than others on the PI/Linux/ArmStrong/Debbian/etc. I’d recommend Logitech since we’ve tested that brand on the PI and it came right up … MS brand seems to work too.
Good luck!
STeven.
I second the PS3 eye. It was hands down the best quality at a higher frame rate of all 5 cameras we tested.
Yea, I’ll be the third to agree with the PS2 eye. Very good fps for the price.
STeven.
I did end up buying the PS3 Eye. Thank you all. It works fine, although I’m yet unsure if it will work in ranges over ~30 feet(I haven’t tested that yet) as the image looks pixalated, although that may be from something on my end.
Where did you get those numbers for FPS? I’m getting a much higher FPS with the Axis 206 Camera(10 - 15 FPS) with the same resolution(and some compression), and I have similar numbers for the PS3.
I’m capturing the image and then doing vision processing for rectangle targets using OpenCV and python. I’m not sending the image anywhere else, which may be what you did.
Although with two cameras(Axis 206 and the PS3 Eye) with vision processing, the fps is hovering around 5, which may be fine. I may reduce the resolution of the PS3 Eye capture image since I’m using it solely for targeting retro-reflective tape less than 90" away(On the pyramid).
Note for anyone using the Raspberry Pi(maybe other similar devices), you may have to bypass certain fuses to power USB cameras. The raspberry pi requires only 700mA to be powered, and so has a 700mA fuse at from the MicroUSB port. You can bypass this by either doing some soldering or by using the GPIO port to purely power your Raspberry Pi; I used a two pin header cable on ports 4 and 6 on the GPIO connected the standard KOP 12 to 5v converter to get power from the PD board. Since the USB camera in my case requires 500mA, and the USB ports too have a polyfuse(fuse that doesn’t permanently trip) limiting current to 140mA, you’ll have to do some soldering to jump those polyfuses for the extra current(and then you may want to super glue the wires you used to jump down). It was a lot of work but the camera works.
http://azrathud.com/images/pi/2-pin-header-small.jpg
2 Pin header in GPIO
http://azrathud.com/images/pi/jumpers-small.jpg
Jumpers over polyfuse
http://azrathud.com/images/pi/all-together-small.jpg
Put together