Study On Methods Of Vedio-Based Face Detection

Face Detection, the critical first step for Automatic Face Recognition System, hasattracted attention of more and more researchers. With the depth of study and theexpansion of application, it's found that, besides Face Recognition, Face Detection alsohas important applications in other fields, such as Video Surveillance at Import andExport, Video Conference, Image Retrieval and Intelligent Human-computer Interaction.Therefore, Face Detection has gradually developed into a relatively independentresearch area and academic direction.After analyzing several important methods of face detection, and according to therequirements for application environment, detection accuracy and real-time of thedetection system, a method is proposed to detect faces in video images at import andexport. With this method, we extract the moving targets from video images at importand export, then carry on face detection on them, based on Color Features and AdaboostAlgorithm. Moreover, we discuss Image Preprocessing in this thesis, such as LightCompensation, Removing Noise by Filter and so on.With collective images from video cameras at import and export, the method offace detection, discussed in this thesis, is evaluated under Matlab. Then the results showthat this method has a high true positive rate, while ensuring real-time and a low falsealarm rate, which satisfies the requirements.