Elaborating on an item previously mentioned among last weekend’s Cornell final projects list, this time with video:
For their ECE final project, [Adam Papamarcos] and [Kerran Flanagan] implemented a real-time video object tracking system centered around an ATmega644 8-bit microcontroller. Their board ingests an NTSC video camera feed, samples frames at a coarse 39×60 pixel resolution (sufficient for simple games), processes the input to recognize objects and then drives a TV output using the OSD display chip from a video camera (this chip also recognizes the horizontal and vertical sync pulses from the input video signal, which the CPU uses to synchronize the digitizing step). Pretty amazing work all around.
Sometimes clever projects online are scant on information…but as this is their final grade, they’ve left no detail to speculation. Along with a great explanation of the system and its specific challenges, there’s complete source code, schematics, a parts list, the whole nine yards. Come on, guys! You’re making the rest of us look bad… Videos after the break…
Basic object tracking: