基于图像处理的铸件缺陷自动识别系统研究

发布时间：2018-03-13 15:21

本文选题：X射线检测　切入点：缺陷检测　出处：《太原科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：X射线无损检测技术在零件缺陷检测中具有广泛应用。但是通过对X射线检测形成的图像进行零件缺陷识别,大都仍然靠人工完成,导致检测系统效率低、检测准确性差。近年来,随着图像处理、模式识别以及计算机技术的发展,采用计算机视觉、机器学习等技术,通过对X射线实现工业零件缺陷的自动检测成为可能,也是目前研究的一个热点,并已有成功应用案例。本文针对精密铸件厂对电动汽车某精密铸件缺陷检测的需求,研究开发了基于X射线检测的铸件缺陷自动识别系统。论文主要在以下几个方面做了研究:1、在深入了解国内外铸件X射线检测的研究现状基础上,利用已有的X射线检测平台,提出了铸件X射线自动检测系统方案,根据要求设计了自动检测系统的硬件和软件。2、针对薄壁铸件的缺陷检测问题,研究了缺陷特征以及特征提取问题,给出了基于图像配准和差影法的缺陷自动识别方法。3、针对X射线成像过程系统固有噪声和随机噪声的干扰问题,比较研究了几种滤波算法的去噪效果,确定了中值滤波作为去噪算法。4、研究了图像配准和差影法的缺陷自动识别算法的实现问题。运用随机Hough变换检测圆形装配孔,圆心作为匹配特征点,完成与模板图像的配准;采用差影法提取铸件缺陷,通过形态学重建将缺陷信息标记在铸件上;开发了相应的缺陷检测软件。5、基于MATLAB软件GUI平台编写人机交互的缺陷自动识别软件,并对开发的软件进行了封装处理。经对零件的测试表明:该检测系统正确识别铸件中的缺陷,可以实现铸件缺陷的自动识别,软件GUI界面也具有较好的实用性。
[Abstract]:X-ray nondestructive testing (NDT) technology is widely used in parts defect detection. In recent years, with the development of image processing, pattern recognition and computer technology, it is possible to use computer vision, machine learning and other techniques to automatically detect defects of industrial parts by X-ray. It is also a hot spot of research and has been successfully applied. This paper aims at the demand of precision casting factory to detect the defect of a precision casting of electric vehicle. The automatic recognition system of casting defects based on X-ray inspection is developed. In this paper, the following aspects are mainly studied: 1. On the basis of deep understanding of the present situation of X-ray inspection of castings at home and abroad, the existing X ray detection platform is used. In this paper, a scheme of automatic X-ray inspection system for castings is proposed. The hardware and software of the system are designed according to the requirements. The defect feature and feature extraction are studied for the defect detection of thin-walled castings. In this paper, an automatic defect recognition method based on image registration and differential image is presented. Aiming at the interference of inherent noise and random noise in X-ray imaging system, several filtering algorithms are compared and studied. The median filter is determined as the denoising algorithm. 4. The realization of automatic defect recognition algorithm based on image registration and difference image is studied. The circular assembly hole is detected by random Hough transform and the center of circle is used as the matching feature point to complete the registration with template image. The defects of castings are extracted by subtraction method, and the defect information is marked on the casting by morphological reconstruction. The corresponding defect detection software .5is developed, and the automatic defect recognition software based on MATLAB software GUI is developed. The software is encapsulated. The testing results show that the system can recognize the defects of castings correctly, and the software GUI interface has good practicability.
【学位授予单位】：太原科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG245;TP391.41

【参考文献】