摘要
圆形标志点检测通常被视为椭圆检测。由于光线、角度等原因,采集到的标志点会出现残缺,另外,当背景或物体较复杂时,非标志点边缘为标志点的提取带来干扰。为此,提出了一种基于椭圆外切矩形性质的圆形标志点检测方法。通过拟合椭圆的准圆来检测椭圆的中心,利用椭圆外切矩形的几何性质来确定椭圆的长短轴位置及旋转角度。为了去除非标志点的干扰边缘影响,进一步构造验证参数,并结合聚类算法,最终提取有效的标志点类。仿真与实物实验表明,该算法拟合精度高,检测性能优良,对具有部分残缺的椭圆有良好的识别效果,且对于复杂情况下的标志点识别仍具有较高的精度和稳健性。
Circular control points detection is usually considered as the detection of ellipses. Due to the illumination, measuring angle and other reasons, the edges of control points may be incomplete. Furthermore, the noise edge of complex background or object will interfere with the extraction of control points. For those reasons, we propose a method of circular control point detection based on the circumscribed rectangle of an ellipse. Firstly, the center position of an ellipse is obtained by fitting director circle. Then the orientation and two semi-axis of an ellipse are determined by using the geometric properties of the external rectangle. The detected ellipses will be verified in order to reduce the detection errors. Finally, a clustering algorithm based method is proposed to extract the efficient control points. The simulation and real experimental results show that the proposed algorithm has high accuracy and excellent detection performance even for incomplete ellipses or cases in complex situation.
出处
《光学学报》
EI
CAS
CSCD
北大核心
2018年第12期246-253,共8页
Acta Optica Sinica
基金
国家自然科学基金(61405034,51475092,51175081)
关键词
机器视觉
椭圆检测
准圆
标志点
聚类
machine version
ellipse detection
director circle
circular control points
clustering
