Real-time tracking for an integrated face recognition system
Stephen James McKenna, Shaogang Gong, H. Liddell · 2007
A real-time tracking system suitable for surveillance applications was implemented on pipeline image processing hardware. Motion estimation was performed using a symmetric spatio-temporal filter in order to determine regions of interest. Normal components of visual motion were obtained at moving `edge' features and Kalman filtering techniques were used for robust object tracking. The use of such a method within an integrated machine vision system for face recognition was discussed. Introduction In order to recognise peoples' faces in realistically unconstrained environments (e.g. such as arise in many security applications), a robust tracking and segmentation is required. This would provide a sequence of roughly segmented face images for recognition or verification purposes. Since people are almost constantly moving, motion estimation provides an e#ective technique for focusing of attention and discarding cluttered, static backgrounds. Perhaps the simplest approach to detecting areas ...