基于灰度曲线交点的结构光编码条纹边缘检测  

Structured Light Encoding Stripe Edge Detection Based on Grey Curve Intersecting Point

作　　者：吴海滨[1] 于晓洋[1] 关丛荣[1] 

机构地区：[1]哈尔滨理工大学,黑龙江哈尔滨150080

出　　处：《光学学报》2008年第6期1085-1090,共6页Acta Optica Sinica

基　　金：国家自然科学基金(60572030);教育部博士学科点专项科研基金(20050214006);黑龙江省教育厅海外学人项目(1055HZ027);哈尔滨市科技攻关计划(2005AA1CG152);黑龙江省研究生创新科研资金(YJSCX2005-238HLJ)资助课题

摘　　要：在分析结构光编码白条纹扩散的原因和实际边缘形态的基础上,提出一种同时检测多边缘并消除扩散影响的亚像素精度方法——交点法。首先向被测表面投射互为反色的两组格雷编码图案,提取两组强度图像中灰度梯度较大的点作为边缘依据点,并对其进行曲线拟合,然后将对应拟合曲线的交点作为边缘。分析了交点法对结构光系统中常见的被测表面倾斜和离焦两个主要误差因素的适应能力。介绍了检测原理及误差因素分析,并进行了边缘检测实验。实验结果表明,对于宽度大于4pixel的编码条纹,边缘检测相对误差小于1%,可消除被测表面倾斜和离焦的影响。Based on analysis of reasons for structured light encoding white-stripe diffusion and practical edge shape,an intersecting point method with sub-pixel precision is presented which can detect multiple edges simultaneously and eliminate the influence of diffusion.Two groups of grey encoding patterns with inverse grey value each other were projected onto the measured surface,and then points with larger grey gradient in the two groups of intensity images were selected as reference edge points.Intersecting points of corresponding curves fitted by reference edge points were treated as edges.The adaptability of intersecting point method to two main error factors such as measured surface sloping and defocus in structured light system was analyzed.The detection principle and error factor analysis are presented.Finally,edge detection experiment was done.Experimental results show,the detection relative error is less than 1% for encoding stripes wider than 4 pixels,and the method can eliminate the influence of measured surface sloping and defocus.

关 键 词：结构光 边缘检测 亚像素 扩散 

分 类 号：TH741[机械工程—光学工程]