多尺度光点图像中心的高精度定位  

High precision location for multi-scale light spot center

作　　者：刘震[1] 尚砚娜[1] 

机构地区：[1]北京航空航天大学精密光机电一体化技术教育部重点实验室,北京100191

出　　处：《光学精密工程》2013年第6期1586-1591,共6页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.51175027);北京市自然科学基金资助项目(No.3132029)

摘　　要：提出了一种可快速、高精度提取不规则光点图像中心的多尺度光点图像中心定位方法。首先,结合图像形态学处理和阈值分割确定光点图像区域,初步确定光点图像大小;然后,计算不同尺度空间下各光点图像区域的Hessain矩阵,由Hessain矩阵特征值确定的判决系数确定最佳尺度,并确定光点图像中心像素级坐标;最后,根据光点区域周边图像二阶泰勒展开式求解光点图像中心亚像素坐标。仿真和实物实验表明,本文算法抗噪声能力强、可实现不同大小光点图像中心的快速高精度定位。在实物试验中,该方法提取光点图像中心的精度优于0.1pixel。目前,该方法已在多套视觉测量系统中得到应用。This paper presented a multi-scale light spot center location method to extract the center lo- cation for irregular light spots in an image fast and accurately. Firstly, the region of the light spot was segmented by morphology operation and thresholding to determine the size of the image with spots. Then, the Hessian matrix was computed in the multi-scale image space in the light spot region. The best scale was determined by a determination coefficient from the eigenvalue of Hessian matrix, and the pixel level center location was computed. Finally, the image intensity function around the spot center pixel location was approximated by second order Taylor expansion and the sub-pixel center lo- cation was computed by finding the maxima of the approximated intensity function. Both the synthetic and real experiments verify that the proposed method has higher location accuracy and stronger ro- bust. In a real experiment, the error of the light spot center location is less than 0.1 pixel. The meth- od proposed has been put into the practice of video measuring systems.

关 键 词：计算机视觉 光点中心 HESSIAN矩阵 形态学 亚像素 定位 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]