Research On Fast Face Detection Algorithm Based On PCA Dimension Reduction

The face detection problem originated from the face recognition, face detection is a key part of the automatic face recognition system, but the early face recognition research is mainly about the strong constraints of face images(in the absence of background image), to assume the position of the face of known or very easy to get to, so face detection problem are not affected by the attention. In the past few years, face recognition has become the most promising biological identity authentication methods,this application background requirements of automatic face recognition system capable of general environmental image has certain ability to adapt, thus facing a series of problems makes face detection began as an independent subject by researchers attention.Firstly, to adapt to the characteristics of Haar high speed and accurate detection of dimension reconstruction is one of the research contents of this paper.Mainly include: the significance of the research of Haar mode feature, feature model in face detection in the process of overlap.Accordingly, we research the reclassification from the aspects of regional pixel feature, which accords with PCA dimension reduction conditions, establish different Haar characteristics of different pixel mode components, PCA dimension.Secondly, Another important content of this paper: to improve the research of intelligent PCA algorithm for adaptive training target. Mainly includes: Research on adaptive high degree of matrix computing algorithm, including feature value, calculation of eigenvector and covariance matrix of the iterative update; the original data set first to obtain the initial results, and then calculate the iterative incremental data influence factor, through the initial node fruit and iterative factor combined with dynamic evolution iteration effect of incremental data sets.This article tries to enhance the efficiency of the algorithm to be easy realized in application, especially accelerate the training speed, the adaptive algorithm to improve, and the practical application of training and detection algorithm in the detection system. In the first place,provide feasible theoretical basis for the application of detection system; secondly,for the experimental detection system provide feasible scheme for the application of reference.