Study On The Method For Face Detection Based On Bionic Intelligence

Detecting human faces robustly and fast is the most important issue during the study offace detection. Combined with bionics and artificial intelligence methods, this thesis studiedon how to improve the robustness and the speed of face detection, some new methods wereproposed in the thesis and listed as follows,1. The speed of face detection is very important during real-time detection for inputimages or input video sequences. A novel face window searching algorithm based onevolutionary agent is proposed which is used to detect faces in gray-scale images. It canquickly find the candidate face windows through the evolutionary computation of distributedagents, which represent different kind of windows. Experimental results show that theevolutionary agent-based searching algorithm can increase the detection speed by 5-7 timescompared with the traditional exhaustive searching method used in some general algorithm.2. A novel face detection method based on bionic pattern recognition is proposed. Facedetection can be regarded as the classification of"face"and"non-face". According to thebasic idea of the continuity of the same kind of samples in feature space, the face detectionalgorithm can be constructed by training face samples which distribute manifold.Experimental results show that the face detection based on bionic pattern recognition has theadvantages of low rejection rate and high detection rate.