Genetic Algorithm-based Face Detection Research

Face detection refers to the process of finding the location, size and gesture of faces in an input image, if there are any faces in it. Face detection is the first step of face recognition, human-computer interface and intelligent vision supervision system. As a key technique in face information processing, face detection has been paid much attention and became a very active research branch in the fields of pattern recognition and computer vision in recent years. In this thesis, a face detection system based on skin colour segmentation, model matching and genetic algorithm is built to detect faces in colour images with complicate background.The main contributions of this thesis are as follows:First, we use the KL-based skin colour model to segment the input colour image and get a rude segmentation image of skin colour regions and background. Then the close operator and the fill operator in morphologic is used to smooth the segmentation image. Finally, the object label operator is used to mark the detached skin colour regions in the segmentation image. On this basis we record the position and area of every skin colour region, and those regions with enough area possession ratio are considered as candidate face regions.Second, we use GA-based model matching to locate faces in the candidate face regions. In the course of GA-based model matching, model matching translates into finding the transformation parameters (including the scaling, rotation and translation parameter), such that the optimal match between the transformed model and the image can be achieved. In this thesis, the average face is used as the model, and the similarity measure includes the modified Hausdorff distance and the geometry feature of human face.Finally, we use Matlab to simulate our face detection system. Experiments show that our system is effective in certain circumstances and can detect faces of different location, size and angle.