基于DSP+FPGA的矿井轨道行人检测系统设计  

Design of pedestrian detection system for underground track based on DSP+FPGA

作　　者：马忠磊[1] 郎文辉[1] 李晓明[1] 卫星[1] MA Zhong-lei;LANG Wen-hui;LI Xiao-ming;WEI Xing(School of Computer and Information, Hefei University of Technology, Hefei 230009, Chin)

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《传感器与微系统》2018年第5期97-100,共4页Transducer and Microsystem Technologies

基　　金：国家国际科技合作专项项目(2014DFB10060)

摘　　要：为了避免矿井机车撞人事故的发生,提出了基于数字信号处理器结合现场可编程门阵列(DSP+FPGA)架构的轨道行人检测系统。通过A/D采样芯片对摄像机输出的模拟信号进行采样,将采集的数字信号由FPGA进行预处理;通过外部存储器接口(EMIF)将预处理后的信号传送至DSP,由DSP执行自适应阈值的边缘检测;基于极角、极径约束的霍夫(Hough)变换,降低维度的梯度方向直方图(HOG)特征提取,结合支持向量机(SVM)行人检测等算法;通过以太网接口,将检测的行人信息传送至上位机显示。实验结果表明:设计的轨道行人检测系统,检测效果良好,帧率可达8 fps,满足矿井环境对安全应用的要求。In order to avoid pedestrian accidents of mine locomotive, a pedestrian detection system based on digital signal processing（DSP） ＋field programmable gate array（FPGA） is proposed. Analog signals outputted from the camera are sampled by A/D chip. Then the FPGA is used to preprocess the collected digital signals. Output signals of FPGA are transmitted to DSP via external memory interface （ EMIF ） port. DSP executes adaptive threshold edge detection;based on polar angle and radius constrain-based Hough transform, reduce histogram of oriented gradient （HOG）feature extraction of dimension, combine SVM pedestrian detection algorithm; detected pedestrian information is transmitted to host computer through Ethernet interface. Experiment show that the proposed orbit pedestrian detection system has a good detection effect, frame rate can be up to 8fps and satisfies the requirement of mine environment for safety application.

关 键 词：数字信号处理器 现场可编程门阵列 霍夫变换 支持向量机 行人检测 

分 类 号：TN751[电子电信—电路与系统]