Study Of Face Detection Algorithm Based On Skin Color Model And Neural Networks

Human face detection is to locate all the faces and determine their sizes in the input image. The subject of face detection was originally put forward as location part. With the development of intelligent information processing, face detection has been paid much attention and become a very active research branch in pattern recognition and computer vision application areas.A detection mode of coarse-to-fine is designed, inclusive of skin color, to detect one or multiple faces in color image. It is a hierarchical approach and combines the skin color rough detection, mathematical morphology hit-miss transform with neural network verification. With the elimination of false areas, the search area will become smaller and smaller, and the detection will be accomplished eventually. The main contributions are as follows:First, on the basis of clustering features of skin color in the chrominance space, the space is selected and the appropriate color model is established for skin color segmentation through the statistics of some sample skin. The candidate face areas can be obtained from a preliminary judgment, reducing the size of image and preparing for the face verification.After rough detection of skin color, the further verification of face is the focus of the study. The candidate face areas are preprocessed and transformed by HMT. Neural Network Algorithm which hides the statistical character of face pattern in its structure and parameter has shown its special advantage in the detection pattern which is complicated and difficult to describe such as faces. on the result of skin color segmentation, a BP artificial neural network is used for the further verification. Through the experiments, accurate classification results and verification of the algorithm in the face target detection and location have shown good performance.Finally, the major face features such as eyes and mouth region are located.