| Face detection is a process to extract the information of face (if there exists) spatial location, size, scale, posture in input image.Face detection technology is a frontier topics in the field of computer vision and pattern recognition. Face detection has important application values in identity authentication; region of interest coding; content-based retrieval; video retrieval; video surveillance; automatic face recognition; intelligent human-computer interaction and many other fileds.This paper starts from face detection algorithm based on AdaBoost which have higher detection performance in present face detection algorithm, aiming to achieve real-time face detection technology in complex background.This paper uses classical AdaBoost algorithm carried out experiment in the training process and testing process in complex background, we found the problems of this algorithm exist in training process and testing process. Both the training process and testing process have the time-consuming problem, the AdaBoost algorithm lost its own advantages and have high false reject rate and missing rate in complex background. Focus on these two issues, this paper proposed optimization method.First of all, aiming at the time-consuming problem in training process respectively optimized the number of haar-like rectangular features and the weak classifier training method, So that reduced the training time to the original one sevens; Secondly, aiming at the time-consuming problem in testing process this paper proposed the approach of combined classifier to reduce detection time while reducing the high false reject rate and missing rate, combined Which classifier and how to combined are the key issues to be resolved in this process, studies shown that add skin color segmentation and placed skin color segmentation before the AdaBoost face detection algorithm will be achieved face detection in complex background which with real-time performance and higher detection rate. |
PDF Full Text Request