| 基于GPU的人脸检测和特征点定位研究 |
| |
| 引用本文： | 张印,董兰芳,王建富.基于GPU的人脸检测和特征点定位研究[J].电子技术,2014(9):38-42. |
| |
| 作者姓名： | 张印 董兰芳 王建富 |
| |
| 作者单位： | 中国科学技术大学计算机科学与技术系 |
| |
| 摘 要： | 人脸分析相关应用越来越广泛,但随着高清视频影像的广泛使用,传统的基于CPU设计实现的程序已难以满足时效性要求。本文基于GPU平台实现了人脸检测和特征点定位的并行化。首先为了加速人脸检测过程,使用Nvidia的CUDA计算范式,通过"窗口级并行"和"分类器级并行"两步实现基于Haar特征的Adaboost算法;然后在人脸检测的基础上,提出一种在常量时间内获得初始模型的方法,并行实现ASM算法。与OpenCV中基于CPU的方法相比,基于GPU的本方法有一定速率提升。
|
| 关 键 词： | GPU CUDA Adaboost算法 ASM算法 高分辨率 |
Face Detection and Feature Localization Based on GPU |
| |
| Zhang Yin,Dong Lanfang,Wang Jianfu.Face Detection and Feature Localization Based on GPU[J].Electronic Technology,2014(9):38-42. |
| |
| Authors: | Zhang Yin Dong Lanfang Wang Jianfu |
| |
| Affiliation: | Zhang Yin Dong Lanfang Wang Jianfu (Dept. of Computer Science and Engineering,University of Science and Technology of China, Hefei 230027, China) |
| |
| Abstract: | Face analysis has been used more and more widely. However, with the wide-spread use of high definition video and images, the traditional program designed and implemented based on has been difficult to meet the requirement for time efficiency. To speed up the process, this paper presents an approach for Adaboost face detection with Haar-like features on the GPU which is regarded as massively parallel coprocessors through Nvidia＇s CUDA computation paradigm by two steps, parallelism processing based on window and classification. Also on the basis of face detection we propose a method for obtaining the initial model in a constant time, parallel implementing ASM algorithm. Compared with the implementation of the method in OpenCV based on CPU, the present method based on GPU achieves certain improvement in speed. |
| |
| Keywords: | GPU CUDA Adaboost algorithm ASM algorithm high-definition |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
