Research on human face location based on Adaboost and convolutional neural network

摘要

To increase speed and accuracy of human face location with Adaboost algorithm, this article proposes a human face location method combining Adaboost with convolutional neural network (CNN). This method firstly uses CNN algorithm to carry out human face training, then applies Gaussian mixture model to establish a background and gets moving targets through background subtraction before making a rough localization by Adaboost in moving targets area and precise location by CNN algorithm. Experimental results show that: this method is superior to Adaboost algorithm in the aspect of human face location speed and accuracy.