Hi guys, I'm looking into the possibility of using Arduino to tell me when an object (any object) moves into the frame from, say, the left edge. I know it's not ideally suited to that sort of processing, but if its able to do it, the cost savings would be substantial it seems. Essentially we'll have a mass of stuff going rapidly from, say, right to left, and slower moving objects going from left to right, or stopping. I would be using an IR capable web cam or some such sensor, possibly only looking at, say, a small subset of the pixels. Has anyone attempted such a thing? Thanks so much 
Just getting a web cam's video into an arduino is a feat in itself. The best you can do is to off load the video acquisition onto an external circuit and then have the arduino access that circuit's memory.
I have not use this device but it says...
Capture low-res video image frames for display or video processing. Give your Arduino the gift of sight!
Perform object detection for computer vision projects.
Yes but it dosn't take a web cam as an input. It is the sort of board I was talking about.
Grumpy_Mike:
Yes but it dosn't take a web cam as an input. It is the sort of board I was talking about.
Well the OP does say
I would be using an IR capable web cam or some such sensor
so assumed the video source was not set in stone.
The Video Experimenter is an amazing bit of kit, but it is low resolution, both spatially and in grey scale, so recognizing "any object" is challenging unless lighting and other image conditions can be very carefully controlled.
Are you after the presence of moving object or are you after the representation.
Where items are to be measured or where items are to be visualized?
sparkdev:
if its able to do it, the cost savings would be substantial it seems.
Compared to what?
Overlay text and graphics onto a video signal from a camera, DVR, DVD player, VCR or any other source of composite video.
Capture low-res video image frames for display or video processing. Give your Arduino the gift of sight!
Perform object detection for computer vision projects.
Over-lay , Putting a Layer over the Image Frame. Not a great feat.
Pal is 25 FRAMES per second, 1 Frame is what it captures, for display (buffered).
Object Detection, 1 Frame per 25 Frames. I don't think you will be capturing anything more than 2 snails fighting 
For this kind of stuff I use a wireless IP cam and open computer vision running on the PC. Control info is then generated on the PC and passed to the arduino.
Even at low res and in B/W an arduino can only no a few frames a min (read cam and store on SD card). With 2k bytes of data space there's really not any way to do real time analysis of an image.
Mark
Pal is 25 FRAMES per second, 1 Frame is what it captures, for display (buffered).
Almost - it actually captures fields.