VideoInput
Goal of the patch
One of the first ideas that popped up during the workshop was to use a microphone and a video camera to track the movements of people on stage, analyze these sounds & movements and transcribe them into datas that could on turn interfere with the light and sound on stage. We gave up the sound analysis to focus on moving image analysis through Pd/Gem. Since we didnt manage to compile the opencv library to track movements inside an environment with moving lights, we simplified both the input and the analysis by building a small model of the stage (well, a cardboard box) with a translucid ground, put a camera underneath, and move a single light source, and use Pd to track it.
Inputs
We first used a Logitech USB webcam wich proved a bad idea, both because of its fixed wide lens,which generates shadow on the edges and aberant optical, and the
- bad exposure corrections, both on Linux as well as on Windows. We havent been
able to use it on OSX because Leopard doesnt support any webcams unless you upgrade. We then used a Sony MiniDV camera which proved great both for its picture quality, exposure correction, and for its zoom which helps greatly to find the right picture position. If you dont have a magic arm, clamps are very useful.
Instead of tracking a finger, which is hard because there were too many shadows, and the Gem object [pix_blob] only tracks one blob at a time and got false positive,we used a USB powered LED on a flexible arm: this both elegant and efficient.
As a side note: we also had to use a black curtain so that the camera doesnt get blinded by the moving light as the beam comes closer to the lens.
Outputs