基于智能相机的铁路异物侵限检测系统设计

发布时间：2018-03-26 11:33

本文选题：异物检测　切入点：图像处理　出处：《北京交通大学》2017年硕士论文

【摘要】：随着我国铁路建设的高速发展,列车行驶速度逐步提高,行车安全问题日趋突出与重要。由于铁路异物侵限造成的事故时有发生,这给铁路的安全运营带来了重大的威胁。现有的基于机器视觉的异物侵限检测方法大多采用嵌入式采集结合PC图像处理的方式,这种方式的异物检测存在实时性差、成本高等问题,针对较长线路的检测时,也无法实现沿线密集式的布置。所以有必要研究一种集图像采集、处理和传输于一体的智能相机及算法,以达到更快速的实时检测处理、报警以及更低廉成本的目的。因此本文提出了一种基于智能相机的铁路异物侵限检测系统,系统针对铁路线路异物设计了基于智能相机的异物侵限检测算法,通过在铁路沿线分布式布置检测节点实现了铁路沿线分布式的异物图像采集、处理、分析、自动报警和异物图像存储的功能。同时,监控中心对本线路各个检测节点的智能相机具有参数配置能力。监控终端可以通过以太网向各个检测节点发送图像处理算法中的配置参数,并可以在线更新,使得异物侵限检测算法能够更快的适应检测环境,提高系统的实用性。论文首先介绍了基于智能相机的铁路异物侵限检测系统的总体方案设计,包括铁路异物侵限检测系统的总体网络结构、结构组成、各模块功能以及总体算法设计。接着详细的论述了智能相机的硬件电路设计、设计依据、主要芯片的选型、选型依据,给出了智能相机的总体电路设计和各主要模块的电路设计。并设计了基于DSP的图像处理算法,利用DSP的硬件特性采用C语言对设计的算法进行编程实现。最终实现了智能相机的图像采集、图像存储、图像传输、图像处理、特征提取、异物报警、以太网传输等基本功能。为了提高异物侵限检测系统的处理速度,本系统分别从程序开发实现、内存分配上进行了深入的研究并较为合理的设计了内存分配体系和算法程序。并从传输方式、存储结构、算法、程序等方面进行了优化。比如本系统提出的单次连通域标记算法,可以在进行连通域标记的同时记录下来多个特征参数,明显的提高了运算效率。为了检验智能相机检测侵限异物的效果,在校园区和铁路现场进行了实验。实验结果表明系统的平均检测帧率达到12帧/秒,满足实时检测的要求。对侵限异物检测报警的准确率达95.83%,满足准确率的要求。
[Abstract]:With the rapid development of railway construction in China, the speed of train running is gradually increasing, and the problem of driving safety is becoming increasingly prominent and important. Accidents caused by the invasion of foreign bodies in railway occur from time to time. This brings a great threat to the safe operation of railway. Most of the existing methods of foreign body intrusion detection based on machine vision are embedded acquisition combined with PC image processing, which has poor real-time performance. The problem of high cost can not realize the dense arrangement along the long line, so it is necessary to study an intelligent camera and algorithm which integrates image acquisition, processing and transmission. In order to achieve faster real-time detection, alarm and lower cost, a railway foreign body intrusion detection system based on intelligent camera is proposed in this paper. The system designed an intelligent camera based detection algorithm for foreign bodies in railway lines. The distributed image collection, processing and analysis of foreign bodies along railway lines are realized by distributing detection nodes along railway lines. Automatic alarm and foreign body image storage functions. At the same time, The monitoring center has the ability to configure the parameters of the intelligent camera of each detection node on the line. The monitoring terminal can send the configuration parameters of the image processing algorithm to each detection node through Ethernet and can update the parameters online. So that the foreign body intrusion detection algorithm can adapt to the detection environment faster and improve the practicability of the system. Firstly, this paper introduces the overall scheme design of the railway foreign body intrusion detection system based on intelligent camera. Including the overall network structure, structure, function and algorithm design of the railway foreign body intrusion detection system. Then, the hardware circuit design, the design basis, the selection of the main chip, the selection basis of the intelligent camera are discussed in detail. The overall circuit design of the intelligent camera and the circuit design of the main modules are given, and the image processing algorithm based on DSP is designed. Using the hardware characteristic of DSP, the algorithm is programmed with C language. Finally, the intelligent camera image acquisition, image storage, image transmission, image processing, feature extraction, foreign body alarm are realized. Ethernet transmission and other basic functions. In order to improve the processing speed of the foreign body intrusion detection system, the system is developed from the program. The memory allocation system and algorithm program are designed in detail, and the transmission mode, storage structure and algorithm are introduced. For example, the single connected domain labeling algorithm proposed by this system can record several characteristic parameters while marking connected domains. In order to test the effect of intelligent camera, experiments were carried out on campus and railway site. The experimental results show that the average frame rate of the system is 12 frames per second. The accuracy of detecting and alarming intrusion foreign bodies is 95.83, which meets the requirements of accuracy.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U298;TP391.41

【参考文献】