Playing with Kinect
I was learning OpenCV for a new art project, but then I learned that the new Microsoft Kinect device had been hacked by a Spanish student after the bounty offered by Adafruit Technologies.
This news changed it all, as Kinect drivers for several platforms were made available. I needed to extract the silhouette of a person from the exhibit space and having a depth image as the one Kinect provides made my job ten times easier. The previous plan was to keep an image of the background and substract the current frame from it to detect changes. But handling changes in lighting conditions was a bit challenging.
What Kinect provides is a regular color (or infrared) webcam plus a second camera that measures depths (ie. each pixel brightness represents the distance to the kinect of that scene point).
If you want to isolate object from an scene that are at a certain distance from the camera, all you need to mark those pixels between a certain value range on the depth image. Simple, fast and highly effective.
Most of the work has been done using OSX and openFrameworks (an open source framework for interactive applications).
Untitled from Miguel Sanchez on Vimeo.