基于图像的缸体零件表面缺陷的高速检测方法

	基于图像的缸体零件表面缺陷的高速检测方法
其他题名	High speed detection method of cylinder block surface defects based on images
	谭亚雄
学位类型	硕士
导师	刘伟
	2015-11
学位授予单位	中国科学院大学
学位专业	机械制造及其自动化
关键词	数字图像处理边缘检测 Canny算子数学形态学制动缸体
摘要	在许多行业中，尤其是那些涉及到人生命安全的地方，总会存在对某些对象进行检测的必要。由此衍生出各种各样的适合于各个行业的检测技术。起初研究的大多数是对某些金属零件表面缺陷检测，这些表面平整并容易用肉眼进行观察，由此也得到了多种检测方法。这里通过与汽车制动主缸缸体主孔检测项目相结合，系统研究了基于机器视觉和数字图像处理技术的零件表面缺陷检测方法。近些年，对这种方法的研究很多，推进了其在各个领域的应用。该方法将采集的被测零件表面图像作为桥梁，通过对采集的图像进行相应的处理来达到检测目的，这种检测技术有着其他方法不具备的一些优点，如在线实时、精度较高、抗干扰能力强等，因此，很有学习和深入研究的必要。本论文的研究内容主要基于以下完成的几项工作：首先，文章阐述了课题研究的背景和意义，介绍了零件表面缺陷的多种检测方法，系统总结了国内外在数字图像缺陷检测技术的研究成果和现状。其次，详细介绍了图像采集实验中的各个仪器设备，提出了一种图像采集方法。在图像采集实验时，遇到了各种技术难点，零件的移动速度控制，相机焦点位置确定，采集窗口大小的选择等，文中给出了相应的解决方法。第三，介绍并对比了多种图像预处理方法，主要有直方图均衡化、中值滤波、均值滤波、小波变换，选取了处理图像效果最好的一种方法；用经典微分算子法对采集的图像进行边缘提取，并比较它们各自处理图像中边缘的完整程度和连续性，最终确定选取最具优势的Canny算子进行边缘检测。第四，阐述了数学形态学在图像处理中的应用，并运用一些合适的数学形态学算法对提取的图像边缘进行边缘优化，提出了一种以间断点为对象进行逐点边缘连接的方法。最后，依据处理后的图像进行缺陷特征提取，给出缺陷特征的判断标准，计算缺陷的周长和面积，再依据周长和面积得到其圆形度，根据这些所得参数对缺陷进行定量分析，以此来完成对零件表面缺陷的检测。实验和处理结果表明，该方法能快速检测出零件是否存在缺陷，并能根据处理结果判断出缺陷的具体特征和所在零件表面的位置，可以满足检测要求。
其他摘要	There will always be a need to detect certain objects in many industries, especially those involving people's life and safety. Thus a variety of detection techniques suitable for all industries were derived. A variety of detection methods had emerged in the detection of surface defects of metal parts. Here, parts surface defect detection method based on machine vision and digital image processing technology were studied systematically by combining with the project which was detecting the main hole of automobile brake master cylinder. In recent years, there was much research on this method, and it had promoted its application in various fields. A method making the collected images of the parts surface as the medium and achieving the goal of detection by corresponding image processing has the advantages which other detection technologies have not such as non-contact, real-time online, the appropriate speed, precision and the scene has strong anti-jamming capability. Therefore, it is very necessary to study and further research. The main research contents of this paper include the following aspects: First, the article expounded the background and significance of the research, introduced the various detection methods of the surface defects of the parts, and summarized the research results and the status quo of the domestic and foreign digital image defect detection technology. Second, this paper introduced the various instruments and equipments in theexperiment of image acquisition in detail, and put forward an image acquisition method. In the experiment of image acquisition, many kinds of technical difficulties, such as the mobile speed control, the determination of the position of the camera, the choice of the size of the acquisition window, and so on, were met. The corresponding solutions were given in this paper. Third, a variety of image preprocessing methods were introduced and compared, which mainly included histogram equalization, median filtering, mean filtering, wavelet transform. And the best method of image processing was selected by using the classical differential operator method for the same image edge extraction and comparing the complete and continuity degree of the edge extracted by each operator. This paper determined choose Canny operator for edge detection eventually. Fourth, this paper described the application of mathematical morphology in image processing, introduces the expansion, corrosion, opening and closing operation, used these algorithms to optimize the edge of the edge extraction image, improved the edge features, and proposed a method of using the point of the object to carry on the edge. Finally, the defect characteristics were extracted from the processed image. The defect feature judgment standard was given. The perimeter and area of the defects were calculated. Then the circular degree was obtained. This paper made quantitative analysis of the defects based on these parameters and completed detecting the parts surface defect. The experiment and actual results showed that this method could detect the parts whether there was defect and the specific characteristics of the defect and its position on the surface of the parts quickly.
语种	中文
文献类型	学位论文
条目标识符	http://ir.ciomp.ac.cn/handle/181722/49238
专题	中科院长春光机所知识产出
推荐引用方式 GB/T 7714	谭亚雄. 基于图像的缸体零件表面缺陷的高速检测方法[D]. 中国科学院大学,2015.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
谭亚雄.pdf（1299KB）	学位论文		开放获取	CC BY-NC-SA	请求全文