基于ZYNQ的人脸检测的快速实现  

Fast Implementation of Face Detection Based on ZYNQ

作　　者：曾爽[1] 应骏[1] 王健[1] 曾维军[1] 

机构地区：[1]上海师范大学数字社区与智能家居研究中心,上海200234

出　　处：《电视技术》2015年第15期144-147,共4页Video Engineering

摘　　要：在ZYNQ平台上,结合加入阈值处理的背景减除法实现了人脸检测的快速实现。ZYNQ是ARM+FPGA的体系架构,利用该芯片的PL(FPGA)端来实现图像的预处理,利用PS(ARM)端对来自PL端的图像利用背景减除法进行前景的提取,然后对前景图片进行人脸检测。最终,实现了人脸检测的时间在ZYNQ平台上平均减少了700 ms以上,而且在识别率方面也有所提高。Combining with the background subtraction method which adds a threshold, fast face detection on ZYNQ platform is im- plemented. ZYNQ has a framework of ARM ＋ FPGA, and in this paper, the PL which is the FPGA on the platform is used to a- chieve pre-processing of the image; the PS which is the ARM on the platform is used to implement background subtraction for the image from PL, then to finish the face detection. Finally,this paper implemenls the reductioo of lace detection time more than 700 ms, and the recognition rate is also improved on ZYNQ.

关 键 词：ZYNQ 人脸检测 背景减除 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]