Since this has been a question I have seen asked a number of times in different forums, I will try to put forth a comparison. Obviously, I'm a little biased, but I'll try to be as objective as possible.
CMUcam2 vs. AVRcam (where C=CMUcam2 and A=AVRcam)
PRICE
C: $199 (assembled...
http://www.seattlerobotics.com)
A: $99 (assembly required, no surface mount components...
http://www.jrobot.net )
SIZE
C: ~2.5"x2.5" (approximate)
A: 2.4" x 1.9"
POWER CONSUMPTION
C: 5V at ~300 mA (anyone know for sure???...number extrapolated from CMUcam)
A: 5V at ~57 mA
EXPANDABILITY
C: Bytecraft C Compiler required for making changes to the software ($1750 USD...
http://microcontrollershop.com/product_ ... cts_id=169 )...not sure of amount of code/data space available...closed source
A: AVR-GCC (free)...~4K of flash, 300 bytes of RAM, and ~450 bytes of EEPROM available...all open source under GPL
CAMERAS SUPPORTED
C: OV6620 or OV7620
A: OV6620
FRAMERATE
C: full frame rate of OV6620 (up to 50 frames/sec)*
A: full frame rate of OV6620 (up to 50 frames/sec)*
*NOTE: There is some ambiguity in how this is measured, since the OV6620 generates interlaced frames back to back, and thus ends up repeating data...but both systems are capable of keeping up with the full frame rate of the camera
NUMBER OF SIMULTANEOUS TRACKED OBJECTS
C: 1
A: 8
NUMBER OF SIMULTANEOUS TRACKED COLORS
C: 1
A: 8
SERIAL BAUD RATES SUPPORTED FOR USER INTERFACE
C: 115.2, 57.6, 38.4, 19.2, 9.6, 4.8 2.4 kbps
A: 115.2 kbps
PC-based USER INTERFACE FOR TESTING
C: Yes
A: Yes
FRAME DIFFERNCING FOR MOTION DETECTION
C: Yes
A: No
COLOR HISTOGRAMMING
C: Yes
A: No
SERVO MOTOR CONTROL
C: Up to 5 servos can be controlled
A: 1 servo can be controlled (requires software modifications, but hardware port is provided)
ANALOG VIDEO OUTPUT
C: Yes (European PAL format for OV6620, NTSC format for OV7620)
A: No
IMAGE DOWN SAMPLING
C: Yes
A: No (not by default)
CONTROL OF THE CAMERA SETTINGS
C: Yes
A: Yes
PROVIDES FRAME DUMP CAPABILITY
C: Yes (uses a FIFO buffer chip to capture a single frame from the camera, and then dumps that image over the serial port...takes about 3 seconds at 115.2 kbps on the OV6620)
A: Yes (no FIFO buffer chip, so it acquires one line at a time from a series of successive frames, dumping each line out the serial port...takes about 4 seconds at 115.2 kbps on the OV6620)
IN-CIRCUIT REPROGRAMMING
C: Requires the SX-Key, available from Parallax and other places for ~$99 (
http://www.parallax.com)
A: Requires an STK200/300 compatiable AVR programmer; schematics are all over the internet, or purchase an assembled one for $14.95 from
http://www.jrobot.net
This provides a reasonably good comparison of the two systems. I guess I should close this off with one final metric:
SUPPORT FROM THE DEVELOPERS IN AN OPEN FORUM
C: No (there is a forum on Acroname's sight, but it is a general vision forum)
A: Yes...you're there right now
I'd certainly welcome any/all feedback or additional comparisons of the two systems. I own a CMUcam (version 1, not 2), and really liked it. There were, however, a number of deficiencies IMHO. This is why I developed the AVRcam. The CMUcam2 does have some benefits, especially the ability to FIFO up an entire image when doing a frame dump. So I guess it comes down to what capability you need.