Exact Eye Localization Based on Adaboost with Haar-like Feature and Independent Components Analysis andIts Embedded System Implementation
| Autor: | KUO, ZHEN-AN, 郭振銨 |
|---|---|
| Rok vydání: | 2016 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 104 In order to effectively raise the accuracy and reliability of face recognition system, especially to effectively raise the accuracy and reliability of eye localization that dominates the performance of face geometric normalization, this paper proposes Exact Eye Localization Based on Adaboost with Haar-like Feature and Independent Components Analysis to localize the eye coordinate to accomplish face geometric normalization. In this proposed exact eye localization method, it makes use of Adaboost with Haar-like Feature to the face region candidate, the eye region candidate, and two irises’ centers. Based on two irises’ centers, the eye region candidate is cropped. Then, it utilizes Independent Components Analysis to extract the features of the eye region candidate and to match the features of the eye region candidate with the ICA-based feature database of 20 eye images for further improvement of exact eye localization and face geometric normalization. Experimental results this proposed exact eye localization method has better accuracy and reliability of eye localization than conventional eye localization methods. On the other hand, this proposed exact eye localization method has been implemented into Android embedded device seamlessly and smoothly. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|