基于FPGA的铁路异物侵限检测硬件平台设计

发布时间：2018-04-22 02:07

本文选题：异物侵限检测 + 图像处理　；参考：《北京交通大学》2015年硕士论文

【摘要】：随着我国铁路建设的快速发展,列车运行速度不断提高,对列车运行安全提出了更高的要求。目前铁路异物侵限事故时有发生,对铁路运行造成巨大的人员伤亡和经济损失。现有铁路异物侵限检测方法存在实时性差等问题,虽然可以检测出异物,但是如果不能及时报警,就不能避免事故发生。为了提高铁路异物侵限检测的实时性,本文提出了一种基于FPGA (Field Programmable Gate Array)的铁路异物侵限检测硬件平台,利用FPGA技术和机器视觉进行图像的实时处理和分析,从而实现铁路沿线异物的实时检测。论文首先介绍了铁路异物侵限检测硬件平台的总体方案设计,包括铁路异物侵限检测系统的总体网络结构、结构组成、各部分功能以及总体算法设计。重点研究了硬件平台图像采集、图像存储、图像处理和图像传输4部分的结构组成和器件选型问题。总体算法由FPGA和ARM两部分组成,FPGA完成异物特征参数提取和初步报警等图像预处理,ARM完成异物的识别分类、跟踪和报警等图像高级处理。接着详细论述了铁路异物侵限检测硬件平台硬件电路的总体构成,给出了各组成单元的电路设计。在电路设计的基础上,实现了硬件平台的图像采集、图像存储、图像帧存控制、主从控制器间通信和PC端模拟服务器等基本功能。为了提高异物侵限检测的处理速度,利用Verilog硬件描述语言移植铁路异物侵限检测算法到FPGA硬件处理器上,实现了背景差分、背景更新、单次扫描连通域标记、异物特征提取等图像处理算法。针对异物识别、分类和跟踪的设计要求,本文提出了单次扫描连通域标记算法,可以在单次扫描期间完成异物多个特征参数(面积、质心、灰度值和外接矩形)的存储,扫描结束后提取参数给ARM。最后,为了验证硬件平台检测异物的效果,在实验室室外进行了异物检测的初步验证实验,在北京北站铁路沿线进行了现场实验。实验结果表明,参数提取正确,处理一帧图像的时间比软件减少50%,速度满足实时检测的要求。由硬件平台系统验证结果可知,系统平均检测频率达到13帧／秒,该处理频率符合实时检测的要求。对于侵限异物、高危侵限趋势异物、列车以及其他噪声具有较好的判别效果,异物侵限检测报警率达到88.57%。
[Abstract]:With the rapid development of railway construction in our country, the speed of train running has been improved continuously, which puts forward higher requirements for the safety of train operation. At present, railway foreign body invasion accidents occur from time to time, causing huge casualties and economic losses to railway operation. The existing detection methods of foreign body invasion in railway have some problems such as poor real-time. Although the foreign body can be detected, if the foreign body can not be alerted in time, the accident can not be avoided. In order to improve the real-time performance of foreign body intrusion detection, a hardware platform for foreign body intrusion detection based on FPGA Field Programmable Gate Arrayis proposed in this paper. The real-time image processing and analysis are carried out by using FPGA technology and machine vision. In order to realize the real-time detection of foreign bodies along the railway line. This paper first introduces the overall scheme design of the hardware platform for foreign body intrusion detection, including the overall network structure, structure, function and algorithm design of the railway foreign body intrusion detection system. In this paper, the structure and device selection of image acquisition, image storage, image processing and image transmission on hardware platform are studied. The whole algorithm is composed of FPGA and ARM to complete image preprocessing, such as extracting characteristic parameters of foreign body and initial alarm, and arm to complete the advanced processing of foreign body recognition, tracking and alarm. Then, the overall structure of the hardware circuit of the hardware platform for foreign body invasion detection is discussed in detail, and the circuit design of each component unit is given. Based on the circuit design, the basic functions of the hardware platform, such as image acquisition, image storage, image frame storage control, master-slave communication and PC analog server, are realized. In order to improve the processing speed of foreign body intrusion detection, the Verilog hardware description language is used to transplant the railway foreign body intrusion detection algorithm to the FPGA hardware processor. The background difference, background update and single scan connected domain mark are realized. Foreign body feature extraction and other image processing algorithms. Aiming at the design requirements of foreign body identification, classification and tracking, a single scan connected domain labeling algorithm is proposed, which can store multiple characteristic parameters (area, centroid, gray value and external rectangle) of foreign body during a single scan. At the end of the scan, the parameters are extracted to ARM. Finally, in order to verify the effect of foreign body detection by hardware platform, a preliminary verification experiment was carried out outside the laboratory, and a field experiment was carried out along the railway line of Beijing North Railway Station. The experimental results show that the parameter extraction is correct, the processing time of a frame image is reduced by 50% than the software, and the speed meets the requirement of real-time detection. The results of hardware platform system verification show that the average detection frequency of the system is up to 13 frames per second, and the processing frequency meets the requirements of real-time detection. For foreign body invasion, high risk trend foreign body, train and other noise has a better discrimination effect, the detection alarm rate of foreign body invasion reaches 88.57.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U298;TP391.41

【参考文献】