Fast Face Detection Method Based On AdaBoost Algorithm

Face detection technology is a subject which can be used to determine whether there were faces in the image and position them. The technology occupies an important position in legal, financial, military, customs, people’s daily life and other fields. According to the difference of human face knowledge, it can be categorized into several types: knowledge-based methods, appearance-based methods, feature-based methods, template matching methods and other methods.In this paper, based on today’s most popular AdaBoost face detection algorithm, the face detection algorithm were trained which used the variance pretreatment and strive to achieve the premise at a relatively high detection rate of the detection rate has a more substantial upgrade, thereby further enhance the real-time application of face detection technology. The main contents include the following aspects:1. Introduces the background and significance of face detection technology, combined with the current research proposes face detection technology and the main difficulties currently exist research directions; a brief introduction of several commonly used internationally recognized face database with face detection algorithm performance evaluation methods and results of this paper are based on the evaluation criteria.2. AdaBoost face detection algorithm based on the analysis and implementation. Introduction section highlights algorithm AdaBoost face detection algorithm theory, introduced Haar-like features, integral image, cascade classifier training also describes AdaBoost AdaBoost algorithm and detection methods and processes based on face detection; achieve some interludes In the algorithm described in the introductory part of the algorithm listed after each implementation method used herein, the two parts together on the face detection process is made very detailed analysis.3. Research on the face detection preprocessing methods. A brief introduction to the most commonly used method of face detection and significance of pretreatment; highlights the variance pretreatment methods used in this article, elaborated from the meaning and purpose of obtaining a simple two-part, determined by the statistics used in this article Face variance threshold.4. Use the face detection algorithm in this paper for Bao database and BioID database and conclusion that: Based on AdaBoost cascade classifier training algorithm is optimum in face detection performance; the variance pretreatment can quickly exclude non-face context to improve the speed of face detection and reduce false detection rate, but there is a negative impact on the detection rate and missing rate.