面向机器视觉测量的液体透镜调焦系统标定方法

Calibration method of liquid lens focusing system for machine vision measurement

高性能光学成像器件是机器视觉测量的重要基础,液体透镜作为一种结构紧凑的快速电控调焦器件,在机器视觉测量领域中具有明确的应用前景。针对液体透镜调焦器件在视觉测量中性能易受环境干扰、系统内参难以准确标定的问题开展研究。首先,构建了基于Optotune液体透镜器件的机器视觉测量系统,分析了液体透镜器件电控调焦工作机理,提出了调焦系统的电流-焦距数学模型,研究了影响液体透镜调焦系统参数的温度、重力因素作用机理,分别提出了温度-焦距、重力-主点位置影响量传递与补偿数学模型;然后,将上述各系统模型与针孔相机成像模型相结合,提出了液体透镜调焦系统的函数化内参表达式,设计了求取表达式全部系数的标定装置与标定流程,并进行了系统内参标定实验,验证了文中内参标定方法的可行性;最后,利用标定所得的内参对边长30 mm的棋盘格靶标进行了测量实验,通过尺寸测量精度反映内参标定精度。实验结果表明,文中方法能够得到比现有插值法更准确的内参标定结果,其中图像角点空间映射位置误差均值为0.10 mm,棋盘格边长测量结果最大相对误差为0.68%,两项误差比插值法所得内参的测量结果分别减小了27.2%和54.4%。

High performance optical imaging device is an important basis for machine vision measurement, as a compact and fast electronic controlled focusing device, liquid lens has a promising application in the field of machine vision measurement. In practical, the performance of liquid lens focusing device is easily interfered by the environment, and the intrinsic parameters are difficult to be calibrated accurately, which limits their application in machine vision measurement. The problems described above were researched, a machine vision measurement system was built based on Optotune’s liquid lens. By analyzing the electronically controlled focusing mechanism of liquid lens, the model between current and focal length was proposed. Then the interaction mechanism between system parameters and temperature factor and gravity factors was studied,followed by the proposition of temperature-focal length model and gravity-main point vertical position model. By combining the system models with pin-hole imaging model, a functional expression of system intrinsic parameters was proposed, and the corresponding calibration device and calibration process to acquire all the coefficients of the expression were designed and verified through experiment. Finally, to evaluate parameter calibration accuracy, measurement test was performed using a checkerboard with sides of 30 mm. Test results show that the proposed calibration method can get more accurate intrinsic parameters, the average positional error of image corner points’ world coordinate mapping result is 0.10 mm, meanwhile the maximum relative error of the length of the side of checkerboard is 0.68%, the errors are reduced by 27.2% and 54.4% respectively compared with test results using intrinsic parameters calibrated by mostly used interpolation method.