摘要
提出一种三步法直线光条图像特征自动提取方法。计算光条各点的HESSIAN矩阵以确定其法线方向,在法线方向利用泰勒级数展开求得光条中心点的亚像素图像坐标,利用光条中心点的空间位置连续性约束,将相邻光条中心点连接成曲线。采用直线近似方法将曲线拆分为独立直线。将位于不同独立直线上的共线点进行融合,构成新的融合直线,通过融合直线长度约束和近似直线的角度约束,分割出单一直线光条的中心点,采用最小二乘直线拟合方法得到结构光直线光条的方程。通过对真实图像的试验,结果表明,提出的三步法直线光条提取方法,能够自动提取出复杂背景中的直线光条,为线结构光视觉测量的现场应用奠定基础。
A three-step extraction method for the image feature of a light stripe in the structuredlight vision is proposed. The sub-pixel center of the light stripe can be obtained with Taylor series expansion in the normal direction which is determined by HESSIAN matrix. The,curve can be linked with the sub-pixel center points of the light stripe subject to the continuity constraint in pixel position. The curve can be split into simple lines with line approximate method. The collinear center points on the different simple lines will be combined to one merged line. And then the equation of the line-stripe can be obtained by the least square line-fitting with the merged center pointsbelong to the same merged line Subject to the length constraint and angle constraint. The experiments conducted on some real structured-tight vision images with complicated background reveal that the proposed three-step extraction method for line-stripe is characterized by high precision, automation and robustness. It advances the structured light vision to be applied widely on site.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2008年第11期215-219,共5页
Journal of Mechanical Engineering
基金
国家自然科学基金(50605002)
教育部新世纪优秀人才支持计划(NC ET-05-0194)资助项目。
关键词
结构光视觉
直线提取
光条
亚像素
Structured light vision Line extraction Light stripe Sub-pixel
