基于嵌入式系统的指针式仪表识别技术的研究-计算机科学与技术；计算机应用技术专业论文.docxVIP

摘要 摘要 I I 摘 要 指针式仪表由于结构简单，使用方便等特点，被广泛应用于电力、化工、 石油、自动化等行业。但是在现实应用的过程中，通过人工判读方法来读取指针 式仪表的显示值，不仅费时费力而且判读过程中误差很大。另外，在恶劣环境下， 由人工操作读取数据比较困难。因此实现指针式仪表读数的自动判读是很有必要 的。实现仪表度数的自动识别具有提高效率、降低生产成本等优点。 随着网络技术和嵌入式技术的不断发展，基于网络和嵌入式技术的远程监控 技术将取代传统的监控方式。本论文围绕以 S3C2410 为核心的硬件平台，采用 嵌入式系统设计方案，使用嵌入式 Linux 为操作系统，以 PC 机为目标显示系统， 将 USB 摄像头采集到的图像，经过网络传输到服务器，完成对仪表图像的检测 任务。 首先研究了嵌入式开发环境的架构，随后研究了嵌入式图像采集平台的框 架，结合嵌入式图像处理的特点，提出了一种改进型 Hough 算法。另外把传统 的差影法与最小二乘法结合，使得图像采集和指针判读以多线程方式并发执行， 这两种方法均适合本系统的要求。 采用机器视觉技术代替传统的人工识别方法，不仅降低了人们的工作强度， 而且也提高了判读的准确度。本论文的研究成果在生产过程的监控和仪表的自动 标定中具有十分重要的应用价值。 关键词：嵌入式系统，图像分割、指针式仪表、Hough 变换、差影法、读数识别 Abs Abstract II II Abstract Analog measurement instruments, because of their simple structure and convenient operation, are applied widely in industry, such as electric-power, chemical engineering, petroleum, automation and so on. But in the process of its application, get the readings always by manual method, which cost a lot of time and sometimes make lots of errors. It is necessary to bring in the technology of automatic recognition point degrees in order to realize automatic calibration, which could advance rate and reduce cost. With the development of network and Embedded ARM technology, traditional remote monitoring system is usurped by a new generation of embedded technologies. In this article, the system is based on the S3c2410 core, uses designed scheme of embedded system, uses Embedded Linux system and is showed target in the pc machine. This system get images through USB camera, transmit them by way of network and complete the recognition task. At first, the framework of embedded development environment is studied, and then the embedded platform of image collect is researched. Combine the features of image processing and produce a new way of point degrees recognitions: “modified Hough transforms”. At last, integrate the “subtraction method” with least square method, which make the system full use of multi-threaded implementation to get the image and recognize point degree