基于Adaboost算法的人脸检测系统实现

Implementation of face detection system based on Adaboost algorithm on embedded system

在嵌入式平台上用软件实现、软硬件协同、全硬件实现三种实现方法,全面考察了Adaboost算法在嵌入式平台的性能。在Xilinx XUPV2P开发板上,软件实现的检测速度是0.5fps,软硬件协同是1fps,而全硬件实现获得了80fps的检测速度,完全达到实时需求。

This paper has studied the performance of Adaboost algorithm on embedded system by comparison of three kinds of method including software, half-hardware and full-hardware. On Xilinx XUPV2P development board, 0.5fps is achieved by the implementation of software method. And, the speed of half-hardware design is lfps. Finally, through full-hardware method, the speed of detection is 80fps which can fulfill the real-time demand.