Real-Time Nose Detection and Tracking Based on AdaBoost and Optical Flow Algorithms

摘要

In this paper we present a fast and robust nose detection and tracking application which runs on a consumer-grade computer with video input from an inexpensive Universal Serial Bus camera. Nose detection is based on the AdaBoost algorithm with Haar-like features. A detailed study was developed to select the positive and negative training samples and the parameters of the detector. Pyramidal Lucas-Kanade optical flow tracking algorithm is applied to the nostrils from a previous nose detection in a frame of a video sequence. Tracking takes 2 ms and is robust to different face positions, backgrounds and illumination. The nose detection and tracking application can be used alone or integrated in a hand-free vision-based Human-Computer Interface.