基于机器视觉的铝铸件表面缺陷检测

发布时间：2018-05-19 05:17

本文选题：机器视觉检测 + 表面缺陷　；参考：《华侨大学》2015年硕士论文

【摘要】：在铝压铸件生产过程中,由于种种原因在铝铸件表面不可避免地会产生气孔、裂纹、划痕等缺陷。目前国内铸造生产车间多采用人工目视检测法,劳动强度大且工作效率低,检测结果易受到人的主观影响;而基于机器视觉的检测技术可以有效克服人工检测方法的上述缺点,具备非接触性、客观性、高效性、精度高等优点,成为未来工业检测的发展方向。本文基于机器视觉和图像处理技术,提出了一套针对铝铸件缺陷的检测识别算法,主要的研究工作包括:(1)在分析铝铸件各类缺陷图像成像特点的基础上,研究如何针对性地将可疑区域都准确地分割出来。特别是对于尺寸微小(直径d?2 mm)的气孔缺陷,如何将其有效地分割出来为铝铸件缺陷图像分割研究的一个难点。对此本文提出基于阈值和形态学相结合的分割方法,可准确完整地分割出气孔缺陷区域,获得其焦点的像素位置。(2)分析研究真缺陷(气孔、裂纹、缩孔、松缩)与伪缺陷(水渍、渣痕、油污)的特征,提出基于区域几何特征、区域亮度特征以及区域中心行灰度曲线波形差异的剔除方法。试验表明,该算法能有效地减少伪缺陷对检测结果的干扰。(3)研究铝铸件表面缺陷的特征表达、提取以及选择。通过分析,本文将铝铸件表面缺陷用几何形状、灰度、不变矩和纹理特征来表达,为了提高识别效率、减少后续分类工作的计算量,采用主成分分析法对特征矢量进行选择降维。(4)研究支持向量机分类器,运用MATLAB平台建立多分类进行仿真实验,在不影响正确率的前提下提高计算效率,降低向量机的复杂程度。
[Abstract]:In the process of aluminum die casting production, defects such as porosity, cracks and scratches will inevitably occur on the surface of aluminum castings due to various reasons. At present, most domestic foundry workshops adopt manual visual inspection method, which is characterized by high labor intensity and low working efficiency, and the testing results are easily affected by human subjectivity, and the detection technology based on machine vision can effectively overcome the above-mentioned shortcomings of the manual inspection method. With the advantages of non-contact, objectivity, high efficiency and high precision, it will become the development direction of industrial detection in the future. Based on machine vision and image processing technology, this paper presents a set of detection and recognition algorithms for aluminum casting defects. The main research work includes: 1) on the basis of analyzing the image imaging characteristics of aluminum casting defects, This paper studies how to segment the suspicious areas accurately. Especially, how to segment the pore defects with tiny diameters (D ~ (2 mm) effectively is a difficult point in the research of aluminum casting defect image segmentation. In this paper, a segmentation method based on the combination of threshold and morphology is proposed, which can segment the stomatal defect area accurately and completely, and obtain the pixel position of its focus to analyze and study true defects (stomata, cracks, shrinkage holes). This paper presents a method based on regional geometric feature, regional luminance feature and the difference of gray curve waveform in the center line of the region based on the characteristics of pseudo-defect (water mark, slag mark, oil stain) and pseudo-defect (water mark, slag mark, oil stain). The experimental results show that the proposed algorithm can effectively reduce the interference of false defects on the detection results. (3) the feature expression, extraction and selection of surface defects of aluminum castings are studied. In this paper, the surface defects of aluminum castings are expressed by geometric shape, gray scale, moment invariant and texture features. In order to improve the recognition efficiency and reduce the computation of subsequent classification work, The support vector machine classifier is studied by using principal component analysis (PCA) to select and reduce the dimension of feature vector. The MATLAB platform is used to establish multiple classifiers to carry out simulation experiments, which can improve the calculation efficiency and reduce the complexity of vector machine without affecting the correct rate.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TG247;TP391.41

【相似文献】