基于贝叶斯模型和数字图像相关的视觉测量  

Vision measurement method based on Bayesian model and digital image correlation

作　　者：谢军昱 许杨剑[1] 王效贵[1] 

机构地区：[1]浙江工业大学机械工程学院,杭州310014

出　　处：《激光技术》2016年第6期866-870,共5页Laser Technology

基　　金：国家自然科学基金资助项目(51175469;51375448)

摘　　要：为了能够快速准确地实现3维测量,提出了一种基于贝叶斯模型和数字图像相关的视觉测量方法。该方法采用数字图像相关法对校正后的图像进行立体匹配,克服了传统立体匹配方法精度不高的问题;采用贝叶斯模型估计图像视差,并将其作为数字图像相关法非线性迭代优化的视差初值,克服粗搜索方法寻找视差初值计算量大、精度低的缺点;基于校正后图像与原图像之间的投影关系,由最小二乘法计算出匹配点的3维坐标。结果表明,基于贝叶斯模型和数字图像相关的视觉测量算法,能够快速准确地实现3维视觉测量。In order to operate 3-D measurement efficiently and accurately, a vision measurement method correlation was proposed based on Bayesian model and digital image. Digital image correlation was used to match the rectified images, which had the ability to improve the accuracy of traditional stereo matching method. The iterative initial disparity of digital image correlation was determined by Bayesian model to overcome the disadvantages of coarse-fine search. Based on the projecting relationship between the rectified image and the raw image, 3-D coordinates of the matched point were then calculated by the least square method. The results show that the proposed vision measurement method based on Bayesian model and digital image correlation has high accuracy and efficiency of 3-D vision measurement.

关 键 词：图像处理 视觉测量 数字图像相关法 贝叶斯模型 立体匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN247[自动化与计算机技术—计算机科学与技术]