机器视觉在仪表监控识别系统中的应用

发布时间：2018-07-08 14:26

本文选题：机器视觉 + 感兴趣区域　；参考：《山东大学》2015年硕士论文

【摘要】：随着国内外智能监控行业和安防产业的发展,机器视觉越来越多地应用在人体行为及表情识别、PCB印刷电路检测、数字和指针仪表识别、产品外观检测、物流物品分类等诸多方面,机器视觉取代人工视觉进入各种领域,大大提高了人们的生产生活效率。在工业领域中,各类仪表如声级计、噪声剂量计、震动测量仪、压力表等在生产中发挥着重大的作用,这些仪表的示值读取通常采用人工方式,效率低、受主观影响比较大,极易产生误读,利用机器视觉来识别仪表示值非常有应用价值。基于上述形势和问题,本论文研究了机器视觉在仪表监控识别系统中的应用。论文首先介绍了机器视觉和仪表监控识别系统的研究意义和发展现状,接着对机器视觉各个组成部分做了详细的介绍和选型方法总结,然后分别介绍了本课题的硬件组成和软件平台,最后通过软件设计完成了对数字和指针式仪表识别的读数识别,并对识别的效果做了总结。通过研究工业应用中的几种数字式和指针式仪表的监控识别,本课题采用了具有优质性能的德国Basler工业摄像头获取图像,借助于微软的MFC、OpenCV、 pylon SDK开发工具实现系统的软件设计。其中本课题的重点是仪表识别的算法设计。系统首先使用机器视觉方法采集仪表的图像,然后利用数字图像处理技术对图像进行预处理操作(使用加权平均法进行灰度化、使用直方图均衡化进行图像增强、进行局部自适应二值化、用开闭操作进行形态学处理)；进而根据数字仪表表盘上的位置、长度、宽高比、轮廓等外观特征信息提取示值区域,对指针仪表使用霍夫变换提取圆形轮廓、定位出指针区域；最后使用图像分割技术将数字和指针分割出来,采用自定义模板匹配方法识别出字符、霍夫线变换检测出指针并计算读数。通过实际场景下的测试,本系统的识别速度和识别准确度均能够达到应用的要求,具有良好的应用价值。本课题所取得的突破和创新部分有如下几点：1.严格按照机器视觉的方法,提出了在仪表监控识别系统中选用照明光源、光学镜头、工业相机的规则或选型指南；2.通过研究和分析Basler工业相机的视频存储格式,成功找到将YUV422格式转换为OpenCV中使用的Mat格式的方法,奠定了使用OpenCV进行数字图像处理的基础；3.设计出一种特征提取方法——基于相对位置、局部长度、宽高比、区域面积的目标轮廓提取方法,实现了感兴趣区域的定位。4.在指针仪表的识别中采用了角度法来读取指针指向位置的示数,即根据指针两端点坐标的连线与水平方向的角度、表盘的最小值和最大值来识别读数。
[Abstract]:With the development of intelligent monitoring industry and security industry at home and abroad, machine vision is more and more used in human body behavior and expression recognition PCB printed circuit detection, digital and pointer instrument recognition, product appearance detection. In many aspects, such as the classification of logistics goods, machine vision replaces artificial vision into various fields, which greatly improves the production and life efficiency of people. In the industrial field, various kinds of instruments, such as sound level meter, noise dosimeter, vibration measuring instrument, pressure gauge and so on, play an important role in production. It is easy to misread, and it is very valuable to use machine vision to identify the indication value of instrument. Based on the above situation and problems, this paper studies the application of machine vision in instrument monitoring and identification system. This paper first introduces the research significance and development status of machine vision and instrument monitoring and identification system, and then makes a detailed introduction to each component of machine vision and summarizes the method of selection. Then the hardware composition and software platform of this subject are introduced respectively. Finally, the recognition of digital and exponential instrument reading is completed by software design, and the effect of recognition is summarized. By studying the monitoring and identification of several digital and exponential instruments in industrial applications, the software design of the system is realized with the help of Microsoft's MFC OpenCVand pylon SDK development tools, and the German Basler industrial camera with excellent performance is used to obtain images. The emphasis of this subject is the algorithm design of instrument recognition. The system first uses machine vision method to collect the image of the instrument, and then uses the digital image processing technology to preprocess the image (using weighted average method for grayscale, histogram equalization for image enhancement, etc. Local adaptive binarization, morphological processing by opening and closing operation), and then extracting the value area according to the position, length, aspect ratio, contour and other appearance information on the digital meter dial. Hough transform is used to extract the circular contour and the pointer region is located. Finally, the numbers and pointers are segmented by image segmentation technique, and the characters are recognized by using the custom template matching method. The Hoff line transform detects the pointer and calculates the reading. The recognition speed and accuracy of the system can meet the requirements of application, and it has good application value. The breakthrough and innovation part of this topic has the following points: 1. In strict accordance with the method of machine vision, the rules or guidelines for selecting lighting source, optical lens and industrial camera in instrument monitoring and identification system are put forward. By studying and analyzing the video storage format of Basler industrial camera, the method of converting YUV422 format to Mat format used in OpenCV has been found successfully, which has laid the foundation of digital image processing using OpenCV. A feature extraction method based on relative position, local length, aspect ratio and area is designed. In the recognition of pointer instrument, the angle method is used to read the indication of pointer pointing position, that is, the minimum value and the maximum value of dial are recognized according to the angle between the coordinates of the two ends of the pointer and the horizontal direction.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41

【参考文献】