摘要
针对大视场下气瓶阀门的自动化操作,提出一种基于对称性投影的三维定位算法。采用级联分类器对阀门进行初定位与分割,通过随机采样一致性(RANSAC)算法提取阀门平面法向,利用规则刚体结构的对称性,以一种轴向投影与自适应阈值相结合的方式获取阀门平面的边缘点对,并结合空间直线方程计算出最终所需的中心坐标,实验结果表明:在690 mm×520 mm的视场范围内,坐标及位姿的均方根误差分别为0.352 mm,0.0145 rad。与常规边缘提取方式相比,该算法能更好地应对噪声与点云的缺失,实现了大视场下阀门中心的精确定位。
Aiming at the automatic operation of a cylinder valve at a wide field of view, a symmetry-based projective algorithm for a three-dimensional(3 D) positioning is proposed in this study. First, the cascaded classifier is used to initially locate and segment the valves. Next, the RANSAC(random sample consensus) algorithm is used to calculate the normal direction of valve plane. Then, we use the symmetry structure of a rigid body to obtain the edge point pairs of valve plane with the combination of axial projection and adaptive threshold. Based on the space linear equation, the center-point coordinate can be thus calculated. Results show that the root mean square errors for coordinates and poses are 0.352 mm and 0.0145 rad at field of view of 690 mm×520 mm, respectively. Unlike the traditional edge extraction algorithm, the proposed algorithm is robust with respect to defective data and noisy data and can obtain precise localization for valve center at a wide field of view.
作者
王潇
刘育梁
李丽艳
Wang Xiao;Liu Yuliang;Li liyan(Optoelectronic System Laboratory,Institute of Semiconductors,Chinese Academy of Sciences,Beijing 100083,China;Center of Materials Science and Optoelectronics Engineering,University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《光学学报》
EI
CAS
CSCD
北大核心
2020年第10期117-126,共10页
Acta Optica Sinica
基金
国家自然科学基金(61875189)。
关键词
机器视觉
三维目标
随机采样一致性算法
图像处理
高精度定位
machine vision
three-dimensional object
random sample consensus algorithm
image processing
high precision positioning
