视觉圆要素检测图像校正技术的研究

Research on Image Correction Technology of Visual Circle Elements Detection

当视觉检测带有多个圆孔要素工件时,由于倾斜角(成像平面与工件平面的夹角)和镜头畸变的影响,孔的圆形特征在图片中突变为椭圆或其他形状,增加了算法复杂度而且降低了检测精度。在分析图像发生突变原因之后,提出了通过角度校正和畸变校正的一种图像校正的方法,使图像能反映真实工件的情况。首先进行角度校正,对找到的亚像素角点进行基于几何距离的椭圆拟合,椭圆短轴与长轴的比即为倾斜角的余弦值,由此将图像校正为倾斜角为零时的图像,角度校正使实际距离相等的点在图像上距离的差值稳定在两个像素之内,从而提高了检测精度;然后进行畸变校正,使用畸变参数及公式对像素点进行重映射。经过实验验证,使用此校正方法校正的圆形要素图像使圆形要素检测算法更容易更准确地检测到圆形要素的特征尺寸。

While visually inspecting workpieces with multiple holes,due to the existence of the slant angle(the intersection angle between the imaging plane and the workpiece plane)and lens distortion,the circular features of the holes mutate into ovals or other shapes in the image,which increases the complexity of the algorithm and reduces the detection accuracy.A kind of image correction method was put forward in which angle and distortion correction was used to reflect the actual workpieces after analyzing the causes of the image’s mutation.Firstly,the angle correction was performed by fitting ellipse based on geometric distances on the found sub-pixel corner points,the ratio of the minor axis to the major axis was the cosine of the slant angle,the new image with zero slant angle was obtained by correcting the image with the slant angle.After angle correction,the difference in the image between the points with the same actual distance was stable within two pixels to improve detection accuracy.Then,distortion correction was performed to remap the pixels using distortion parameters and formulas.Experimental verification shows that using the corrected image,the circular features can be detected more easily and accurately.