Abstract:This article began with the machine vision development process starting, first introduced the concept of face detection. Then, after the introduction of AdaBoost algorithm, the paper described in detail on the the calculation of points chart, eigenvalues and the strong classifier specific training process. After the scientific analysis of the efficiency in the training, the article focused on reducing error rates based on the contribution of feature selection methods to reduce the rubbish features of the system resource consumption and adverse effects. Finally, the experimental results of face detection and a conclusion were reached.