摘要
基于视觉的监测系统在野外进行长期工作时,安装了监测系统的平台自身会发生6-自由度(6-DOF)的微小晃动,从而给测量结果带来较大的误差。提出了一种线性求解相机微小位姿变化的方法:通过构造垂直双相机模型,引入垂直固连约束关系,可线性解算出相机6-DOF的微小变化。仿真验证表明:当相机发生三个欧拉角不超过30′,三个平移向量不超过10 mm的微小位姿变化时,与典型的基于单目透视n点(PnP)的位姿估计算法相比,所提算法在解算精度、鲁棒性以及运行效率上都具有较大优势;室外实验表明:所提方法与单目PnP算法对比能更优地补偿观测平台自身微小运动引起的测量误差,在6个控制点下测量44 m处目标点的沉降量,利用所提方法修正平台位姿变化后绝对测量误差的平均值降至0.2 mm,验证了所提方法的有效性和实用性。
The observation stage,on which the vision-based monitoring system is installed,experiences slight 6-degree-of-freedom(6-DOF)shaking when working for a long time.As a result,it brings great errors to the measurement results.Therefore,a linear solution method for micro-motion variations of the cameras was proposed.Micro 6-DOF pose variations of the cameras were then linearly calculated with a newly constructed model of two orthogonal cameras in fixed connection.Simulation verification shows that when the three Euler angles of the cameras are less than 30′and the three translation vectors are less than 10 mm,the proposed method delivers better solution accuracy,robustness,and computational efficiency than the typical monocular perspective-n-point(PnP)pose estimation algorithm.Meanwhile,outdoor experiments show that the proposed method better calibrates the measurement error caused by the micro-motion of the observation stage.The subsidence of the target point 44 m away from the observation stage was measured and the result shows that in the case of 6 control points,the average absolute measurement error after the pose variation of the stage is calibrated decreases to 0.2 mm,which verifies the effectiveness and practicability of the proposed method.
作者
陈文均
尹义贺
张跃强
刘肖琳
于起峰
Chen Wenjun;Yin Yihe;Zhang Yueqiang;Liu Xiaolin;Yu Qifeng(Institute of Intelligent Optical Measurement and Detection,Shenzhen University,Shenzhen,Guangdong 518000,China;College of Physics and Optoelectronic Engineering,Shenzhen University,Shenzhen,Guangdong 518000,China)
出处
《光学学报》
EI
CAS
CSCD
北大核心
2021年第23期175-185,共11页
Acta Optica Sinica
基金
国家重点研发计划(2019YFC1511102)
国家自然科学基金(12002215)。
关键词
机器视觉
摄像测量
位姿估计
微动平台
垂直双相机
machine vision
videometrics
pose estimation
micro-motion stage
dual-orthogonal-camera