Create a basic framework for tracking objects in a camera between frames. Use a really simple 2D (2.5D with size info) Kalman filter to fill in lost tracks. Do automatic tracking re-initialization. This would be something close to the predator tracker.
The code would look something like this.
tracker = Tracker(featureExtractor, feature_description /e.g. color/, camera, optional params)
while(1):
if(tracker.hasTrack()):
print tracker.position()
Anonymous