指针式仪表读数自动识别系统研究
发布时间：2018-08-25 13:26
【摘要】：由于指针式仪表结构简单,而且还具有防水、防尘、防电磁干扰等优点,所以在石油工业领域中,指针式压力表有着很广泛的应用。在指针式仪表读数识别中目前主要还是以人眼观测为主,为了提高检测的快速性、准确性和高效性,本文针对石油工业中常用的指针式压力表,设计了一个自动识别压力值的系统。该系统主要是通过机器视觉,利用图像处理技术自动得出压力表示值。本系统采用500万像素的CCD相机,先将采集到的仪表图像预处理,主要包括三个步骤:首先通过中值滤波来去除表盘图像中的椒盐噪声,同时也能很好地保存图像边缘部分;然后采用分段线性变换来增强指针和刻度线等重要感兴趣部分的信息;最后利用Canny边缘检测仪表图像的边缘,根据表盘的形状和位置等特征提取出仪表表盘的边缘,并通过该边缘拟合出的圆分割出含有指针和刻度线的表盘轮廓。通过预处理得到了感兴趣的表盘图像,然后就只需确定指针和刻度线的位置关系即可确定压力表示值。由于本文研究的耐震压力表的刻度线前后分布是不均匀的,并且仪表里面的阻尼液遮挡了部分刻度线,传统的角度法或者是距离法不适用本系统。本文提出了图像归一化算法,通过改进的平面二维插补算法,限制插补搜索的起始和终点角度,去除表盘图像中的文字标识并将圆形表盘图像转化成矩形图像。然后通过指针处于不同位置处的两幅仪表图像制作出无指针的刻度线表盘模板,并将其拼接处融合并平滑,将该表盘模板利用zhang细化算法处理得到单像素的刻度线模板,以便刻度线精确定位。将单像素表盘模板通过改进的霍夫直线变换检测出刻度线,并根据刻度线位置关系补全出被遮挡的刻度线,这样即得到了完整的刻度线模板,同时将每个刻度线的横坐标值记录并保存到数组中。最后将采集到的仪表图像与刻度线模板图像通过差影法得出指针的位置,依据前后段刻度的不同分度值与指针和刻度线之间相对应的位置关系,利用距离法最终得出指针式仪表的读数。本系统最终将所有识别算法集成到MFC应用程序,并根据指针式仪表校验规则,该程序采用模块化设计,主要包括三个模块:注册模块、识别模块和分析记录模块。通过图像处理算法识别的压力表示值与数字精密仪表读数的比较分析,生成压力表的校验报告。
[Abstract]:Because of its simple structure and advantages of waterproof, dustproof and electromagnetic interference, it is widely used in the field of petroleum industry. In order to improve the speed, accuracy and efficiency of the detection, this paper aims at the pressure gauge commonly used in the petroleum industry, which is mainly based on the human eye observation in the recognition of the reading of the needle meter, in order to improve the speed, accuracy and efficiency of the detection. An automatic pressure recognition system is designed. The system is mainly through machine vision, using image processing technology to automatically get the pressure gauge value. In this system, 5 million pixel CCD camera is used to preprocess the collected instrument image, which includes three steps: first, the salt and pepper noise in the dial image is removed by median filter, and the edge part of the image can be saved well; Then we use piecewise linear transformation to enhance the information of important parts of interest, such as pointer and scale line. Finally, we use Canny edge detection to detect the edge of the instrument image, and extract the edge of the meter dial according to the shape and position of the dial. The contour of dial with pointer and scale line is segmented by the circle fitted by the edge. The interested dial image is obtained by preprocessing, and then the pressure gauge value can be determined only by determining the position relationship between the pointer and the dial line. Because the distribution of the scale line of the seismic pressure gauge studied in this paper is not uniform, and the damping liquid in the instrument blocks part of the scale line, the traditional angle method or the distance method is not suitable for this system. An image normalization algorithm is proposed in this paper. The improved planar two-dimensional interpolation algorithm limits the starting and ending angles of interpolation search, removes the character marks in the dial image and converts the circular dial image into a rectangular image. Then, the dial template without pointer is made by two instrument images at different positions, and the splicing place is fused and smoothed. The dial template is processed by zhang thinning algorithm to get the scale line template of single pixel. So that the precise positioning of the dial line. The single pixel dial template is detected by improved Hoff line transformation, and the occluded scale is completed according to the position relation of the scale line, thus the complete scale template is obtained. At the same time, the horizontal coordinate values of each scale are recorded and saved to the array. Finally, the position of the pointer is obtained by using the difference shadow method between the collected instrument image and the scale template image, and the corresponding position relationship between the pointer and the scale line is obtained according to the different indexing values of the front and rear segments of the scale. The distance method is used to get the reading of the pointer meter. Finally, the system integrates all the recognition algorithms into the MFC application program, and according to the verification rules of the exponential instrument, the program is designed by modularization, which includes three modules: registration module, identification module and analysis record module. The calibration report of the pressure gauge is generated through the comparison and analysis of the pressure representation value recognized by the image processing algorithm and the digital precision meter reading.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
[Abstract]:Because of its simple structure and advantages of waterproof, dustproof and electromagnetic interference, it is widely used in the field of petroleum industry. In order to improve the speed, accuracy and efficiency of the detection, this paper aims at the pressure gauge commonly used in the petroleum industry, which is mainly based on the human eye observation in the recognition of the reading of the needle meter, in order to improve the speed, accuracy and efficiency of the detection. An automatic pressure recognition system is designed. The system is mainly through machine vision, using image processing technology to automatically get the pressure gauge value. In this system, 5 million pixel CCD camera is used to preprocess the collected instrument image, which includes three steps: first, the salt and pepper noise in the dial image is removed by median filter, and the edge part of the image can be saved well; Then we use piecewise linear transformation to enhance the information of important parts of interest, such as pointer and scale line. Finally, we use Canny edge detection to detect the edge of the instrument image, and extract the edge of the meter dial according to the shape and position of the dial. The contour of dial with pointer and scale line is segmented by the circle fitted by the edge. The interested dial image is obtained by preprocessing, and then the pressure gauge value can be determined only by determining the position relationship between the pointer and the dial line. Because the distribution of the scale line of the seismic pressure gauge studied in this paper is not uniform, and the damping liquid in the instrument blocks part of the scale line, the traditional angle method or the distance method is not suitable for this system. An image normalization algorithm is proposed in this paper. The improved planar two-dimensional interpolation algorithm limits the starting and ending angles of interpolation search, removes the character marks in the dial image and converts the circular dial image into a rectangular image. Then, the dial template without pointer is made by two instrument images at different positions, and the splicing place is fused and smoothed. The dial template is processed by zhang thinning algorithm to get the scale line template of single pixel. So that the precise positioning of the dial line. The single pixel dial template is detected by improved Hoff line transformation, and the occluded scale is completed according to the position relation of the scale line, thus the complete scale template is obtained. At the same time, the horizontal coordinate values of each scale are recorded and saved to the array. Finally, the position of the pointer is obtained by using the difference shadow method between the collected instrument image and the scale template image, and the corresponding position relationship between the pointer and the scale line is obtained according to the different indexing values of the front and rear segments of the scale. The distance method is used to get the reading of the pointer meter. Finally, the system integrates all the recognition algorithms into the MFC application program, and according to the verification rules of the exponential instrument, the program is designed by modularization, which includes three modules: registration module, identification module and analysis record module. The calibration report of the pressure gauge is generated through the comparison and analysis of the pressure representation value recognized by the image processing algorithm and the digital precision meter reading.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关期刊论文 前10条
1 蒋超;牛宏侠;;基于改进Radon变换的直线钢轨识别算法[J];铁道标准设计;2017年04期
2 张翠芳;杨国为;岳明明;;Zhang并行细化算法的改进[J];信息技术与信息化;2016年06期
3 李媛媛;黄泉源;侯智新;;基于霍夫变换的OSTU改进算法在路面裂缝检测中的研究[J];电子设计工程;2016年05期
4 陈怀玉;杨e，
本文编号：2203011
本文链接：https://www.wllwen.com/shoufeilunwen/xixikjs/2203011.html
