基于Zynq的LBP人脸识别算法的实现  

Realization of LBP Face Recognition Algorithm Based on Zynq Computer Engineering and Applications

作　　者：彭凌霄[1] 符意德[1] 

机构地区：[1]南京理工大学计算机科学与工程学院,江苏南京210094

出　　处：《自动化技术与应用》2016年第12期100-103,共4页Techniques of Automation and Applications

摘　　要：为了提高人脸识别算法的处理速度,采用FPGA作为主要的运算器件,在Zynq芯片上实现了基于LBP算子的人脸识别算法。对LBP算法进行分析,并在原算法的基础上提出改进。结合Zynq的特征,对算法的实现进行软硬件划分,将计算量较大的特征提取和直方图统计用FPGA进行硬件加速,对算法的流程控制和计算量较小的特征比对用ARM实现。运行的结果表明,算法能够有效的进行人脸识别,通过改进算法和利用VIVADO HLS对算法的优化设计可以进一步提高资源利用率并提高算法效率。The face recognition algorithm based on LBP operator is realized on Zynq chip using FPGA as the main coumputing device. After analyzing the LBP algorithm, and on the basis of the original algorithm, improving is put forward. Combined with the features of Zynq, the algorithm is divided into the part of hardware and software, the larger amount of computation of feature extraction and histogram statistics are accelerated with FPGA. The flow control algorithm and comparison among the characteristics of a small amount of calculation is realized with ARM. The results show that the algorithm can realize face recognition effectively. By improving the algorithm and the optimization of the algorithm using VIVADO HLS can further improve the utilization rate of resources and improve the efficiency of the algorithm.

关 键 词：人脸识别 Zynq芯片 软硬件协同设计 现场可编程门阵列 局部二值模式算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]