Fast method for face location and tracking by distributed behaviour-based agents

摘要

The authors present a novel approach for fast human face location and tracking in colour image sequences. First, a number of colour-sensitive behaviour-based agents are uniformly distributed in the 2-D image environment to search the skin-like pixels and mark the face-like regions by activating their evolutionary behaviour. Then the shape information of each face candidate region is examined and the lace regions are finally determined by fuzzy shape feature analysis. Experimental results show that the tracking speed is about 30 frames per second for 160 × 120 video size and 12 fps for 320 × 240 video size.