【作者】 李学聪；

【导师】 汪仁煌；

【作者基本信息】 广东工业大学 ， 控制理论与控制工程， 2014， 博士

【摘要】 精密指针表读数直观、结构简单、使用方便且性价比高,在生产实践中得到了广泛的应用。为了确保生产质量及生产安全,仪表需定期进行检定,确保其使用精度。因精密指针表在读数时有特殊要求,所以人工检定过程耗时长,容易引入误差从而影响检定精度。若能够在误差范围内实现指针表读数的自动识别系统,将大大提高全国各计量单位指针表的检定效率和检定精度,因此实现指针表读数准确自动识别具有重要的理论意义和应用价值。指针表读数自动识别经过多年的研究,基础理论和应用都有了一定的发展,工业用低精度指针表读数自动识别己进入实用阶段；而目前尚未查到精密指针电工仪表(不低于0.5级)读数自动识别系统的实际应用。每年有数以万计的精密指针表需要计量单位进行检定,由于这类指针表的图像采集、图像预处理、图像分割、读数识别和误差修正等关键技术还没有得到很好的解决,因此该类指针表的检定工作目前只能全部由人工来完成。本文以0.5级毫安指针仪表为对象,从仪表成像开始分析,对光照成像系统、图像预处理、刻度线图像分割、刻度线特征提取、刻度线检测拟合、读数指针获取、读数识别方法和误差修正等问题进行了系统的研究,提高了精密指针表读数的识别速度和识别精度。本文主要的研究成果及创新点简述如下：1.根据精密指针表的物理特性和误差要求,构建了方形无影光源侧面投射的成像系统,解决了以往只能在黑箱里进行图像采集的难题。提出了一种快速获取仪表读数区域图像的方法,为快速识别读数提供了基础。2.针对读数区域图像刻度线特征的亮度分布,提出了改进的耦合冲击复扩散滤波算法和流形上改进的热导方程滤波算法,并分别对读数区域图像进行降噪仿真实验,取得较好的降噪效果。耦合冲击复扩散滤波算法在运算速度和去噪效果方面都优于流形上的热导方程滤波算法,能够明显增强图像中的刻度线图像特征,为准确提取刻度线特征图像提供了基础。3.根据刻度线图像的形状及其密集分布的特点,提出了改进的Snake模型和改进的变分水平集的两种图像分割算法对刻度线特征进行分割提取,两种算法各有优缺点,可根据需要选用。基于Snake模型的分割算法精度较高,但需要人工参与；基于变分水平集的分割方法无需人工参与,能快速提取多刻度线特征图像,更适用于细长类多目标的图像分割。4.分析精密指针表的结构特点,提出了精密指针表单目成像时引入的视点误差的修正方法,实验结果达到预期的精度要求；构建了双目精密指针表成像模型,推导了视点误差的修正公式,为实现精密指针表全自动读数识别提供了理论基础。5.提出了一种基于虚拟表盘的判读法,结合基于刻度圆心的刻度线拟合算法,消除了偏心误差,给出了减小刻度线分布非线性误差的方法,整体上提高了读数识别精度。本文结合研究成果,给出了精密指针表读数自动识别系统实现必须满足的要求,并进行了仿真实验,取得良好的实验结果,为研制精密指针表自动检定系统的设计提供理论依据,具有一定的实用价值。更多还原

【Abstract】 Precise pointer meter is used largely because it is easy to use and reading intuitively with simple structure and cost effective. For safety and quality of the production the pointer meter must be regular verification, to ensure its practical precision. Because precise pointer meter reading needs special requirements, so its artificial verification is time-consuming and easy to produce error and affecting the calibration accuracy. If the readings automatic identification of pointer meter can be implemented within the allowed error, Pointer Meter Automatic Verification System (PMAVS) will greatly improve the verification efficiency and verification precision of pointer meter at metrology organizations. So Implementation of the pointer meter readings accurate automatic identification has important theoretical significance and application value.The research of pointer meter reading automatic recognition has some definite development for several years, low precision pointer meter readings automatic recognition has been realized; Precise Pointer Meter (not less than0.5) Automatic Recognition System isn’t currently found. Every year tens of thousands of precise pointer meter need verification. Because the big scale area, the many scale lines, high accuracy of reading recognition and the computational complexity of the image processing, it is not implemented that the readings automatic identification of precise pointer meter, so all verification of them must be done by artificial work.This dissertation object of study is milliampere precise pointer meter of0.5accuracy grade. The main studied problems of this dissertation include:lighting imaging system of pointer meter, image preprocessing, scale pointer image segmentation, scale line feature extraction, scale line detection and fitting, reading recognition method and error correction. Those were studied systematically to solve the problem of reading recognition speed and accuracy. The main research results and innovation points of this dissertation can be expressed as follows:1. According to the physical characteristics and error requirements of precise pointer meter, it is builded the side projection imaging system with square shadowless light, to solve the problem of image acquisition in the black box only. This dissertation proposes a fast method to obtain the reading area image of meter, providing a basis for fast identification of reading.2. According to brightness distribution characteristics of the scale line image in the reading area, the denoising experiment of reading area image is filtered by both filtering algorithms. A filtering algorithm is the coupling of shock filtering and complex diffusion filter algorithm. Another filter algorithm is the manifolds filtering algorithm based on heat equation theory. The denoising experiment results of the former are superior to the latter in computing speed and denoising effect. The former can significantly enhance the image of scale line image characteristics, providing a basis for the accurate extraction of scale lines feature image.3. According to the characteristics of the scale lines image, it is proposed the image segmentation algorithm based on Snake model and variational level set for the scale lines image experiment simulation. Both methods have their advantages and disadvantages which can be chosen depending on your needs. The segmentation accuracy of former method is higher, but needs artificial participation. The latter without artificial participation, can rapidly and accurately extract image multiple scale lines characteristics. The latter method is suitable for long and thin kind of multi-objective extraction.4. Analyzed the structure characteristics of Precise Pointer Meter, the method of the monocular imaging model was proposed to correct the viewpoint error, and the simulation experimental results have obtain the desired accuracy requirements. Building the binocular imaging model of precise pointer meter, the viewpoint error correction formula is deduced, providing a theoretical basis foundation in order to realize precise pointer meter automatic reading recognition.5. A new reading recognition methods based on virtual dial was proposed, combining with the scale line fitting method based on the center of scale to eliminate the eccentricity error. The method is given to reduce the nonlinear error of scale line distribution, improving the reading recognition accuracy.Based on the research results, this dissertation gives the requirements of the precise pointer meter reading automatic identification system, and has carried on the simulation experiment, obtaining the good results and providing a theoretical basis foundation to develop precise pointer meter automatic verification system. This dissertation has a certain practical value.更多还原