Face Detection using Haar-like Features and the AdaBoost Algorithm in a Cascade of Classifiers
| Autor: | Chen-yu Lin, 林宸宇 |
|---|---|
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 97 Biometrics includes recognition of fingerprinting, retina, iris, palm-print, face, and so on. Face recognition has been widely accepted due to its non-intrusive characteristics. Face detection is the most important pre-processing step, and its goal is to identify human faces in an image. Therefore, the important issue is to detect faces in real-time and with very high detection rate. In 2001, Viola & Jones introduced a fast face detection system which uses Haar-like features, AdaBoost algorithm, and a cascade detector that can achieve high detection rate and low false positive rate. After that, many researchers have improved Viola & Jones’s method to obtain higher detection rate. Since the number of Haar-like features proposed by Viola & Jones is over 180,000, a lot of computation is required during the training phase. In this study, we propose a strategy to select Haar-like features and reduce the computation burden to one-third of the original. We also propose a strategy to design the cascade detector and to reduce false positive rate. Even if fewer features are used in our cascade detector, the detection rate is over 80%. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|