摘要
随着行人检测技术的发展和应用,迫切需要能够进行实时处理的嵌入式行人检测系统。采用ZYNQ-7000作为算法平台,设计一种基于HOG与AdaBoost级联分类器的行人检测系统。利用FPGA的并行特性,采用流水线结构替代传统的串行结构,实现HOG算法加速;将AdaBoost级联分类器保存在FPGA的BRAM中,通过查找表的方式,在单个时钟周期内即可完成匹配判断。利用ZYNQ的软硬件协同设计,根据功能和资源进行软硬件分工,提高系统性能。实验结果表明,该设计方法在保持同等检测性能的条件下,检测速度相比ARM片上系统提高了44倍。
With the in-depth research and extensive application of human detection in mobile devices,there is an urgent need for a real time embedded human detection system.Such a system was proposed based on HOG algorithm and AdaBoost classification using ZYNQ-7000 as the hardware platform.Considering the FPGA’s parallel feature,the pipeline structure was adopted instead of the traditional serial structure to speed up the HOG algorithm.AdaBoost classifiers were stored in the BRAM of FPGA and the matching process required only a single clock cycle using looking up table method.A hardware and software co-design was employed in this implementation according to the functions and resources,which greatly improved the performance of the system.Experimental results show that the proposed design is 44 times faster than the ARM-based design under the same detection performance.
作者
嵇达龙
张尤赛
王亚军
JI Da-long;ZHANG You-sai;WANG Ya-jun(School of Electronics and Information,Jiangsu University of Science and Technology,Zhenjiang 212003,China)
出处
《计算机工程与设计》
北大核心
2020年第1期238-245,共8页
Computer Engineering and Design
基金
国家自然科学基金面上基金项目(61371114)
