面向实时三维扫描的交叉激光中心自适应精确提取算法  

Accurate adaptive extracting algorithm of cross laser stripe center for real-time 3D scanning

作　　者：姚海滨[1] 翁金平 程筱胜[2] 崔海华[2] 

机构地区：[1]扬州市职业大学机械工程学院,江苏扬州225009 [2]南京航空航天大学机电学院,江苏南京210016

出　　处：《南京理工大学学报》2016年第6期708-714,共7页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(51305201);江苏省自然科学基金(BK20130793);江苏省博士后科学基金(1301104C);江苏省精密与微细制造技术重点实验室开放课题

摘　　要：针对激光扫描的普适性问题,提出1种基于自适应梯度阈值和Harrise角点跟踪的交叉激光中心提取方法。由梯度图像直方图自适应确定二值化阈值,以此提取激光边缘并计算激光条初始位置。结合Hessian矩阵计算初始中心的亚像素中心,提取图像中交叉激光的交叉点。利用Harrise角点跟踪算法提取实时三维扫描过程中的交叉点,并以此分割出2条激光线,完成交叉激光对应的物空间重建。实验结果证明该文算法具有较高的鲁棒性,误差小于0.05像素,经GPU加速处理,单张图像提取时间小于2 ms。A cross laser stripe center extracting algorithm is proposed based on an adaptive gradient threshold and Harrise comer point tracking algorithm for the universality of laser scanning. A binary threshold is determined adaptively using a gradient image histogram, and the laser edge is found and the initial position of stripes is calculated. The sub-pixel positions of the cross laser stripe center are calculated using the Hessian matrix, and the cross point of the cross laser stripes is extracted. The cross point of real-time three dimensional （3D） scanning is extracted using Harrise corner point tracking algorithm and two laser stripes are separated, and the object space of the cross laser stripes is rebuilt. Experimental results show that this algorithm is robust;the error of this algorithm is less than 0.05 pixel;it takes 2 ms to process one frame by GPU acceleration.

关 键 词：三维扫描 交叉激光 激光扫描 梯度阈值 梯度图像 直方图 中心提取 

分 类 号：O438[机械工程—光学工程]