基于机器视觉的圆形工件二维尺寸测量方法研究

发布时间：2018-02-27 19:19

本文关键词： 机器视觉图像处理尺寸测量亚像素精度人工鱼群算法　出处：《安徽工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：在工业生产中,为了保证产品的质量,必须对产品进行诸如尺寸、表面光洁度以及强度等参数的检测。随着企业生产规模化,传统的人工检测方法会因检测工作量变大、检测速度和精度提高等影响,导致检测质量波动较大。基于机器视觉的检测技术由于具有自动化程度高、精确性好,并且可以实现非接触测量等优点,被广泛地应用于工业检测和质量控制领域。为了顺应现代工业检测的发展趋势,本文利用机器视觉与图像处理技术,在现有的实验条件下以圆形工件的直径测量作为研究对象,对基于机器视觉的圆形工件二维尺寸测量方法开展了深入的研究和分析。概括本文所做的研究工作,主要包括以下几点:1)硬件平台搭建。在分析系统构成以及各个硬件性能参数的基础上,从实际需求出发,选择合适的硬件,搭建出包括计算机、CCD图像传感器、镜头等在内的测量平台。同时,对测量系统的软件结构和工作流程也做了简要地介绍。2)图像预处理方法分析与选择。对常用的图像滤波、阈值分割、边缘检测等算法进行研究,通过试验比较它们对工件图像的处理效果,找到适合本文研究的工件图像处理和检测方法。3)提出了一种基于人工鱼群算法(AFSA)的圆形工件检测方法,以实现通过较少的内存资源和时间消耗,快速而准确地进行圆形工件检测。首先,从边缘图像上随机选取非共线的三个边缘点表示一个候选圆,即人工鱼个体;然后,统计位于该候选圆上的边缘点个数来计算个体的适应度,在适应度值的指导下,人工鱼通过多次觅食、聚群、追尾等行为最终找到组成圆的三个边缘点的最佳集合;最后,通过与现有的基本Hough变换圆检测、随机Hough变换(RHT)圆检测、基于遗传算法(GA)的圆检测等检测方法的比较,可以看到,我们所提出的方法能够从复杂图像和含噪图像中准确的进行圆检测,同时具有较强的鲁棒性、较快的速度和亚像素精度。4)完成系统的标定。使用该标定系统对圆形工件进行大量的测量实验,并对测量结果进行分析,总结产生误差的原因和消除误差的办法。研究结果显示,本文的测量方法能够满足一般工业检测的精度要求,也为机器视觉在工业测量中的具体应用提供了一种途径。
[Abstract]:In order to ensure the quality of products in industrial production, the parameters such as size, surface finish and strength must be tested. Because of the high degree of automation and accuracy, the detection technology based on machine vision has the advantages of high automation, good accuracy, and can realize non-contact measurement, etc. It is widely used in the field of industrial inspection and quality control. In order to conform to the development trend of modern industrial inspection, this paper uses machine vision and image processing technology to measure the diameter of circular workpiece under existing experimental conditions. In this paper, the two-dimensional dimension measurement method of circular workpiece based on machine vision is studied and analyzed, and the research work done in this paper is summarized. On the basis of analyzing the composition of the system and the performance parameters of each hardware, according to the actual demand, select the appropriate hardware, and build the CCD image sensor including the computer. At the same time, the software structure and workflow of the measurement system are also briefly introduced. 2) the image preprocessing methods are analyzed and selected. The common algorithms of image filtering, threshold segmentation, edge detection and so on are studied. By comparing the processing effect of the workpiece images with experiments, the method of workpiece image processing and detection suitable for this paper is found. 3) A circular workpiece detection method based on artificial fish swarm algorithm (AFSA) is proposed. In order to detect circular jobs quickly and accurately through less memory resources and time consumption. Firstly, three edge points of non-collinear edges are randomly selected from edge images to represent a candidate circle, that is, artificial fish individuals. Statistics the number of edge points on the candidate circle to calculate the fitness of individuals. Under the guidance of fitness, artificial fish finally find the best set of three edge points that make up the circle through many foraging, clustering and rear-end behaviors. By comparing with the existing basic Hough transform circle detection, random Hough transform circle detection, genetic algorithm based circle detection and so on, we can see, The proposed method can detect circles accurately from complex images and noisy images, and has strong robustness. The system is calibrated with high speed and sub-pixel precision. The calibration system is used to carry out a large number of measurement experiments on the circular workpiece, and the measurement results are analyzed. The research results show that the measurement method in this paper can meet the precision requirements of general industrial detection and provide a way for the concrete application of machine vision in industrial measurement.
【学位授予单位】：安徽工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】