基于机器视觉的工件尺寸检测系统研究

发布时间：2018-01-12 04:13

本文关键词：基于机器视觉的工件尺寸检测系统研究　出处：《长春工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：尺寸测量是获得数据的重要方法,也是衡量产品是否合格的重要技术手段。传统的尺寸测量方法是人工手动测量,其耗时费力且精度不高。随着技术的发展,手动测量精度及效率已经不适应工业自动化,尤其是高精度产品的要求。随着机器视觉的发展,视觉检测技术得到了不断发展,它具有精度高、速度快、非接触等优点,在工业现场得到了广泛应用,具有广阔的发展前景。本文以齿轮为研究对象,以机器视觉基本原理和数字图像处理技术为基础,从提高齿轮数据检测精度入手,对齿轮检测系统搭建及采集以及软件设计、图像预处理技术、边缘检测技术、亚像素细分、相机标定进行了详细分析,实现了齿轮数据的精确测量。主要研究内容有:首先,完成了机器视觉工件尺寸检测系统硬件和软件设计。按照齿轮尺寸检测的具体要求,简要介绍了硬件的性能,同时进行了设备选型,在实验室环境中搭建了所用的检测平台,并且获得了齿轮图像。在软件方面,利用MATLAB图形用户界面设计工具设计齿轮图像处理界面,并详细介绍了该界面的设计流程及处理过程。其次,对采集到的齿轮图像进行预处理。采集到的图像不可避免的含有不同程度的噪声,对图像进行滤波处理以得到有利于边缘检测的图像。针对齿轮图像二值化分割方面,利用OTSU法获取阈值,进而对图像进行分割。然后,齿轮图像边缘检测及亚像素细分算法研究。针对边缘检测方面,通过对经典边缘检测算法分析,并利用标准Lena图像进行实验对比,选用Canny算子提取齿轮图像的边缘。边缘检测算法得到的精度较低,利用亚像素边缘检测算法细分边缘,从而提高边缘定位精度。采用Zernike矩亚像素算法进行边缘细分,然后优化了一些数据,通过两组实验验证改进算法的有效性。实验结果表明,改进算法能够更加有效地检测出图像的边缘,提高了定位精度。最后,摄像机标定和尺寸测量结果及误差分析。利用标定板得到了计算齿轮尺寸所需要的标定系数。利用搭建好的硬件平台和设计好的软件系统完成齿轮图像的处理过程,得到齿轮尺寸数据。详细分析了可能影响尺寸检测精度的误差因素。本文设计的基于机器视觉的工件尺寸检测系统,能够精确检测出齿轮的各个尺寸数据信息。通过实验验证,本文算法能够满足检测精度以及速度的要求,是一种有效的检测算法。该系统能够提高工业现场的检测效率,具有一定的理论和实用价值。
[Abstract]:Dimension measurement is not only an important method to obtain data, but also an important technical means to measure whether the product is qualified. The traditional measurement method is manual measurement, which is time-consuming and low precision. With the development of technology. Manual measurement accuracy and efficiency have not been adapted to the requirements of industrial automation, especially high-precision products. With the development of machine vision, visual inspection technology has been continuously developed, it has high accuracy and fast speed. Non-contact has been widely used in industrial field and has a broad prospect. This paper takes gear as the research object, based on the basic principles of machine vision and digital image processing technology. In order to improve the accuracy of gear data detection, this paper analyzes in detail the construction and acquisition of gear detection system and software design, image preprocessing technology, edge detection technology, sub-pixel subdivision, camera calibration. The main research contents are as follows: firstly, the hardware and software design of the machine vision workpiece size detection system is completed, according to the specific requirements of gear size detection. This paper briefly introduces the performance of hardware, at the same time carries on the equipment selection, sets up the testing platform in the laboratory environment, and obtains the gear image. In the aspect of software. The gear image processing interface is designed by using the MATLAB graphical user interface design tool, and the design flow and processing process of the interface are introduced in detail. Secondly. The gear image is preprocessed. The image inevitably contains different degrees of noise. The image is filtered to get the image which is favorable for edge detection. In the aspect of binary segmentation of gear image, the threshold is obtained by OTSU method, and then the image is segmented. Then the image is segmented. Gear image edge detection and sub-pixel subdivision algorithm research. For edge detection, through the analysis of classical edge detection algorithm, and use standard Lena image for experimental comparison. Using Canny operator to extract the edge of gear image, the accuracy of edge detection algorithm is low, and sub-pixel edge detection algorithm is used to subdivide the edge. In order to improve the accuracy of edge location, the Zernike moment sub-pixel algorithm is used to subdivide the edge, and then some data are optimized. The effectiveness of the improved algorithm is verified by two groups of experiments. The experimental results show that the improved algorithm is effective. The improved algorithm can detect the edge of the image more effectively and improve the positioning accuracy. Finally. Camera calibration and dimension measurement results and error analysis. Calibration coefficients needed to calculate gear dimensions are obtained by using calibration plate. Gear image processing is accomplished by using built hardware platform and designed software system. Cheng. The gear dimension data are obtained and the error factors that may affect the accuracy of the dimension measurement are analyzed in detail. A workpiece dimension detection system based on machine vision is designed in this paper. Can accurately detect the gear dimensions of the data information. Through experimental verification, the algorithm can meet the detection accuracy and speed requirements. The system can improve the detection efficiency of industrial field and has certain theoretical and practical value.
【学位授予单位】：长春工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TG86

【相似文献】