摘要
将大模板高斯递归实现引入到结构光条纹中心提取中,提出了一种基于感兴趣区域(ROI)的结构光条纹中心混合图像处理方法。结合图像的阈值化和膨胀算法,自动分割出结构光条所在区域作为光条提取的ROI,利用高斯卷积递归实现获得ROI内光条纹各点的Hessian矩阵,并确定光条纹各点的法线方向,最后在法线方向利用泰勒级数展开求得ROI内光条纹中心的亚像素图像坐标。实验表明,基于ROI的结构光条纹中心混合图像处理方法具有精度高、鲁棒性好和自动化程度高等特点,所提出的算法大大地减少了结构光条纹提取的冗余计算,实现了光条纹中心线的快速高精度提取。在保证光条提取的精度和鲁棒性前提下,所提出的算法将光条提取速度提高了10多倍,为结构光视觉三维测量的实时应用奠定了基础。
Recursive implement algorithm of Gaussian convolution with tremendous template size has been applied to the cen ter extrication of structured light stripe. A composite image processing method to detect the sub-pixel center of structured light stripe based on region-of-interest（ROD is proposed. By combining image threshold with image dilation,ROIs of structured-light are automatically segmented in a measured image. The normal directions of light stripe in ROI is determined by Hessian matrix, which is obtained from recursive implement of Gaussian convolution. The sub-pixel center of the light stripes can be found in normal directions with Taylor series expansion. Experiments show that the proposed method drastically reduces the redundancy computation and implements high-accurate center extrication of structured light stripe with high speed.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2008年第11期1534-1537,共4页
Journal of Optoelectronics·Laser
基金
国家自然科学基金资助项目(50605002)
教育部新世纪优秀人才支持计划资助项目(NCET-05-0194)
航空科学基金资助项目(05151062)
