Multi-pose head tracking using colour and edge features fuzzy aggregation for driver assistant system

Abstract

providing a fast, reliable head tracking module, is a confirmed requirement in human-computer interface (E.g. intelligent vehicles and camera mouses etc.). In this paper we develop a robust, computationally low cost, head detection and tracking system. Preferred features are skin colour and edges which serve to complement each other and tend to binary masks. Then these masks are converted to gray-level masks. Fuzzy Inference Systems (FIS) can efficiently combine these masks to segment current frame of video. Video frame rate of designed system is 4 f/s and is implemented in a driver face analysis context to find position and head pose. The system is used in a constrained environment with a static background and can work in various illuminations. Experimental results show the algorithm's robustness in the cases of extreme movement and large degree rotations. It should be mentioned that the soft computing idea which we introduce here can be implemented in other object's tracking applications.