计算机与现代化 ›› 2010, Vol. 1 ›› Issue (8): 143-146.doi: 10.3969/j.issn.1006-2475.2010.08.040

• 图像处理 • 上一篇    下一篇

基于AdaBoost算法的人脸检测在嵌入式系统中的实现

李先锋1,花小朋1,赵 峰2,孔令东1   

  1. 1.盐城工学院信息工程学院,江苏 盐城 224051; 2.上海交通大学微电子学院,上海 200240
  • 收稿日期:2010-03-20 修回日期:1900-01-01 出版日期:2010-08-27 发布日期:2010-08-27

Implementation of Face Detection Based on AdaBoost Algorithm in Embedded System

LI Xian-feng1, HUA Xiao-peng1, ZHAO Feng2, KONG Ling-dong1   

  1. 1.School of Information Engineering, Yancheng Institute of Technolog, Yancheng 224051, China;2.School of Microelectronics, Shanghai Jiaotong University, Shanghai 200240, China
  • Received:2010-03-20 Revised:1900-01-01 Online:2010-08-27 Published:2010-08-27

摘要: 为解决人脸检测实时性问题,针对AdaBoost算法纯软件实现的瓶颈,提出基于FPGA平台的硬件加速策略,采用流水线处理技术实现积分图像的快速计算。实验使用PowerPC405处理器VirtexTMII Pro平台FPGA,在输入图像大小为352×288像素的条件下,检测速度达到每秒50帧,检测率为98%,误检率约1%,实现了实时人脸检测的要求。

关键词: 人脸检测, AdaBoost算法, 硬件加速, 现场可编程门阵列

Abstract: To solve the technical problem of realtime face detection and break through the bottleneck of AdaBoost algorithm, a hardware acceleration strategy based on FPGA is used to calculate the integral squared image by pipeline architecture. The system can detect a face in 352×288 image at a speed of 50 fps on PowerPC405 VirtexTMII Pro FPGA platform, and achieve a detection rate of 98% with a false alarm rate of 1%, so it can meet the requirements of realtime face detection.

Key words: face detection, AdaBoost algorithm, hardware acceleration, FieldProgrammable Gate Array(FPGA)