I started playing around with OpenCV recently and came across a Haar Classifier code from Intel. After writing a script to process off a webcam in real-time, I noticed quite a bit of false positives. To solve this, I added a smoothing filter. The smoothing filter will lag slightly behind the face. When a face passes through the smoothing filter, I put a green box around it, and then store the image stream of a close-up on the person's face. The smoothing filter will do some basic face tracking, which enables me to get high quality images of a face to feed into a soon to come facial recognition system. I will be posting a demo of this. However, the system is currently locked up in a room in the engineering building (I was playing with hooking it up to servos and actually tracking the person).