Research And Realization Of Face Real Time Detection Technology

Face detection refers to the process of detecting face from a static image ordynamic video and extracted facial features. It is the first step of face recognition andanalysis and has important application in many fields, such as authentication, tracking,security access control, intelligent human-computer interface, and so on. Face detectionis a complex, challenging research topic because it is affected by image background,brightness and human posture and other factors In recent years, face detection hasbecame an independent project which is pay more attention by many scholars.The research progress of face detection and typical face detection methods areanalyzed in this thesis. Combined with the actual situation, the good real-time AdaBoostalgorithm is chosen as the core algorithm. In this thesis, thereare three contents asfollows:①The process original PCA model to boosting algorithm is introduced. Theboosting algorithm is improved to form the AdaBoost algorithm. Based on the analysisof AdaBoost algorithm shows that as long as there are more weak classifiers enough, thearbitrary detect precision can be achieved.②The principle of AdaBoost algorithm is analyzed in detail. First collection oftraining samples are collected and preprocessed, such as normalization, grayscaleconversion, image equalization, which makes them meet the requirements of trainingsamples. Then, the training methods of weak classifier are discussed According to thetraining error minimum standards, T optimal weak classifiers are selected in many weakclassifiers, and according to the way of weighted voting, the T optimal weak classifiersare boosted into several strong classifiers, so that the strong classifier can satisfy adetection rate. The strong classifiers are cascaded to form the cascade classifier. Thestrong classifier that is in the front of cascade classifier can get rid of most of thenon-human face window. Thus, the detection efficiency is improved. Finally, theapplication of AdaBoost algorithm for face detection principle is elaborated. throughanalysis and comparison between equal proportion enlargement the detection windowand equal scaled down to the detection window, the method of equal proportionenlargement the detection window is used to improve the detection efficiency.③A real-time face detection system is researched and realizd. OpenCV, MFC andthe cascaded classifier are combined to design and implement a real-time face detection system which can real-time detect and display face. Based on the static image anddynamic image detection experiments show that face detection rate can reach90%, theaverage detection time is65ms, and the system can satisfy the requirement of real-timedetection for multiple attitude and face.