基于形态学的人工标志中心亚像素定位方法研究

Center subpixel location of artificial target based on morphology

在使用圆形人工标志的计算机视觉检测中,标志的定位精度直接影响到检测的整体精度。以往的中心定位算法对人工标志的中心定位精度差、速度慢,已不能满足精密、快速的检测要求。在经典数学形态学的基础上提出采用改进的扩展形态学的方法提取圆形人工标志的边缘,通过最小二乘拟合方法计算人工标志中心。该方法具有精度高和速度快的优点。实验表明其定位精度在0.02-0.03像素之间,能够满足高精度计算机视觉检测的要求。

During the computer vision detection process of round artificial targets, the precision of artificial targets＇ centers strongly influences the precision of the whole project. Therefore, a number of approaches are proposed to determine artificial targets＇ centers. However, almost all of these approaches are not precise and quick enough to detect the artificial targets＇ centers, and thus could not satisfy the requirements of precise and rapid detection. A new method of edge detection based on extended mathematical morphology is adopted to extract the edges of round artifidal targets. Experimental results are encouraging; the locating precision is within 0.02 to 0.03 pixels both for the simulated image and real image and is comparable with VSTARS software and this method is fairly valuable for precise and rapid computer vision detection technology.