摘要
为了提升3D显示的用户体验,提出了一种基于人脸检测式的3D显示器。针对Haar特征Adaboost人脸检测算法在视频处理中实时性差的特点,提出了一种针对视频的改进人脸检测算法,使系统中的人脸检测速度获得大幅提升。同时使用了高效摄像机标定方法,保证了在人脸检测中的坐标转换精确度。此外,研究了投影变换和3D模型绘制的过程,并针对DSP平台进行了性能优化。实验结果表明了系统的整体实时性性能和良好的视觉体验。
To enhance the user visual experience, a system of 3D display based on face detection is proposed. For the real-time need of face detection in DSP platform, a novel method of face detection in video processing based on Adaboost algorithm is proposed. The efficient camera calibration method is used to ensure the human face detection accuracy of the coordinate transformation. In addition, the projection transformation and the 3D model information optimize the performance in the DSP platform. Finally, the real time and visual performance of the system is validated by the practical experiments.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第11期3750-3754,3758,共6页
Computer Engineering and Design
基金
浙江省科技厅重点实验室建设基金项目(2008E10004)
关键词
人脸检测
HAAR特征
摄像机标定
针孔模型
投影变换
face detection
Haar-like features
camera calibration
pinhole model
projection transformation
