Face Detection Based On Adaboost

Face detection is the detection of the existence of human face in the input imageor video sequences and the process in which extracting information from the data contained human face. The information usually refers to the size, shape posture and position of human face. The detection of human face is one of the important topics in the fields of computer vision, pattern recognition and artificial intelligence. It is not only the prerequisite of face recognition but also the basis of many classification problems. Consequently it is significant in the application of identification, security, video retrieve and surveillance.Based on the Gentle Adaboost, a new face detection system based on multi-threadand multi-feature is designed and realized in this paper.Firstly,a detailed description of Adaboost based on Haar-Like feature and the detect result such as the long time of training, the high false accepted rate is shown by a largeof experiments.Secondly, the paper employ the MB-LBP feature in the human face detection, shortened form of Muti-Black LBP, which is actually the expansion of LBP operator and makes it possible to analyze image integrally. This has great significance in the human face feature presentation. Moreover, it involves Adaboost algorithm to train and detect. In the processing, we conclude that the algorithm involved both MB-LBP feature and Adaboost algorithm has shorter training time,lower false rate and longer detection time. And we come to the conclusion that the detection based on MB-LBP featurehas better effect than the one based on Harr-Like feature, but lower detection speed.Finally,we proposed a new algorithm,according to the Haar-Like feature and MB-LBP respectively,in the purpose of reducing the training and testing time.This algorithm is on the basis of multi-thread and multi-feature.In a lage of experiments,it is confirmed that this algorithm is effective in both reducing the rate of false detections andspeeding of training and detection.