装配位姿视觉检测系统中红外靶点质心定位  

Centroid Location Algorithms of Infrared Target in Computer-vision-based Assembly Pose Detection System

作　　者：杨博文[1] 张丽艳[1] 叶南[1] 冯新星[1] 李铁林[1] 

机构地区：[1]南京航空航天大学,南京210016

出　　处：《中国机械工程》2014年第7期965-971,共7页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(50875130);国防技术基础科研项目(J152011C001)

摘　　要：结合装配位姿视觉检测系统的研发,为提高红外LED靶点质心亚像素定位精度与稳定性,对红外LED靶点图像的灰度分布模型进行了研究,提出了一种基于自由曲面拟合的质心定位算法以获取靶点图像的亚像素中心。根据靶点图像的灰度分布,该算法通过双三次样条插值生成靶点灰度分布曲面,利用拟牛顿法求取曲面顶点即靶点图像中心的亚像素坐标。实测实验显示该算法在图像噪声水平较低的高精度测量环境中,当靶点距离摄像机约5 m和12 m时,与高斯曲面拟合法和加权灰度重心法相比,该算法都能更准确地描述靶点的灰度分布,生成的灰度曲面更加接近靶点图像的真实分布,在高精度测量环境中具有更高的测量精度和稳定性。For the development of computer-vision-based assembly pose detection system, in or- der to improve the precision and stability of centroid location of infrared LED target, a gray distribu- tion model of target image was researched, and a new method for estimating the sub-pixel location of infrared LED target based on free surface fitting algorithm was presented, and by bi-cubic spline inter- polation according to the gray distribution of target image, the proposed method generated target gray surface and evaluated the sub-pixel location which was the vertex of target gray surface by Quasi-New- ton method. Real-data experimental results indicate that in high precision measurement with low noise, the free surface is close to real target gray distribution, the centroid detection has higher preci- sion and stability in high precision measurement, comparing with Gaussian centroid method and weighted gray centroid method, when the targets are at about 5 m and 12 m distance from the camera.

关 键 词：图像处理 光学测量 亚像元 质心定位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]