Improved AdaBoost Face Detection Algorithm Based On Fusion Of AFSA And PSO Optimization

Face detection is mainly through the analysis of existing in the input image of face segmentation and extraction processing,getting all the face,such as location,size,and that is the key technology in the field of visual recognition in machine vision,video monitoring and security systems and has a very extensive application.With the highly development of social informatization,image and video has become a common way of storing information about society,huge numbers of information,however,how to efficiently in the huge data needed to detect the characteristic information,has become a very urgent research subject.In the process of face detection technology to explore many more mature method is proposed,but all have certain conditions,the AdaBoost algorithm in the application of the technology has a certain expected effect.But from the point of view of the fact that,just according to some features of face samples provided by the identification effect is not ideal.Aiming at the shortcomings of the traditional AdaBoost algorithm,this article has carried on the related improvement work,specific work content is as follows:(1)This paper learned and analysed the current face detection technologies,summarized their existing problems.Research mainly focus on the AdaBoost algorithm for research and improvement,introduced its characteristics,performance and application in the field of face detection in detail analysis the related principle and detection performance of AdaBoost face detection algorithm.(2)AdaBoost face detection algorithm has a problem which is weight coefficient easy to fall into local optimum in the process of iteration,In order to make up for the defect,this paper absorbed PSO algorithm's advantages which convergence speed is fast,combined with the best optimization property of the artificial fish algorithm to optimize PSO algorithm further.thus improved PSO algorithm owns two advantages,they are global optimization and random search.then applied this two advantages in AdaBoost face detection algorithm.through this method,we can make weak classifier weighting coefficient to find the optimal value,obtain the optimal combination of the weak classifier coefficient,thus can improve the detection precision rate.At the same time,the improved PSO algorithm can effectively avoid the face feature exhaustive search,improve the local search ability,speed up the training speed of weak classifier,greatly reduce the training time.(3)For a picture,its length and width both are 24 px,there are about Haar-like features need to calculate,as the AdaBoost algorithm is established on the basis of the statistical model,it needs a lot of training set samples,so we need more characteristic quantity calculation,thus training time will certainly be affected.In the training framework of AdaBoost algorithm,due to the existence of all kinds of face features,existing Haar-like feature model can not meet the demand of real-time detection.So this paper based on the most obvious features of face area to extend Haar-like features,here we choose eyes and mouth,the purpose of extending Haar-like features lies in exclusion of lower Haar-like face sample features,thus shorten the training cycle.In conclusion,this paper based on traditional Ada Boost algorithm,presents two improved algorithms and verified by simulation experiments.The experimental results show that the application of improved PSO optimization algorithm to the new development of AdaBoost algorithm framework,detection performance is effectively,improved algorithm not only has important research significance,especially has extensive practical application value.