Face detection and posture recognition in a real time tracking system

The main purposes of this paper are to achieve human face detection and head posture recognition, as well as to track a dynamic image in real time via camera. First, skin-color region is detected. After morphological operations, unnecessary noise is removed, and the method of seed region growing is used to mark pixel blocks. Then the skin-color region is determined whether or not each block is a human face. If it is not human face, it is discarded. Otherwise, wavelet transform is used to decompose the face image. A low-frequency sub-band face image is captured by wavelet transform, and two-dimensional principle component analysis (2DPCA) is used to recognize head posture. Face color histograms are used to build face models, and faces are traced by the self-organizing hierarchical particle swarm optimizer with time-varying acceleration coefficients (HPSO-TVAC) algorithm. In order to solve the face masking problem, adaptive seeking windows are applied. When a human face is not detected, a large seeking window will be used, which will zoom in or out depending on the best global fitness.