Human Face Detection Based on Improved AdaBoost Algorithm
Yi Zhang · Jisuanji fangzhen · 2011
Recently,AdaBoost training algorithm has been widely used in target detection and recognition,as well as many other pattern classification fields.Focusing on the problem of excessive sample weight growth and feature redundancy in the construction of human face detector by conventional AdaBoost algorithm,this paper proposes an improved human face detection approach based on adaptive sample weight updating rule and genetic algorithm.First,both False Negative Rate(FNR) and False Positive Rate(FPR) are taken into the step of sample weight renewing,which feedbacks the result of classifier into itself in order to control the classifier structure efficiently.Then the genetic algorithm with strong search ability is adopted to optimize those selected features and their parameters to build a system that can search out most human faces in images with lower FPR and less weaker classifiers.Simulations show that compared with the conventional AdaBoost algorithm,the proposed algorithm can effectively avoid sample weight distortion,eliminate feature redundancy and reduce false alarm rate while maintaining a high detection rate,achieving a higher detection speed with much more accuracy.