弱特征表面形貌测量的低成本渐进式匹配方法  

Low-cost incremental registration method for measuring the surface topography of weak features

作　　者：房长帅 刘赵阳 王倩雯 张效栋[1,2] FANG Changshuai;LIU Zhaoyang;WANG Qianwen;ZHANG Xiaodong(State Key Laboratory of Precision Measurement Technology and Instruments,Tianjin University,Tianjin 300072,China;School of Precision Instruments and Optoelectronic Engineering,Tianjin University,Tianjin 300072,China;Standard Optics technology Tianjin Co.,Ltd.,Tianjin 300072,China)

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072 [2]天津大学精密仪器与光电子工程学院,天津300072 [3]三代光学科技(天津)有限公司,天津300072

出　　处：《红外与激光工程》2024年第7期169-178,共10页Infrared and Laser Engineering

基　　金：国家自然科学基金项目(62373274);天津市技术创新引导项目(基金)(23YDTPJC00640)。

摘　　要：弱特征连续表面的匹配问题是计算机视觉和图像处理中的一个挑战性问题,基于标准夹具的特征辅助可以准确地实现测量数据与模型的匹配,但该方法成本较高。针对该问题,提出了一种低成本的渐进式的匹配算法,首先基于带有边界罚函数的“点到面”ICP算法实现粗配准,接着通过对测量数据进行两步微调即可对标工业常用评价结果。以汽车玻璃的面形误差评价为例,仿真和实验结果表明,基于所提方法匹配后的面形误差接近传感器本身的误差级别。对于40 cm×40 cm的汽车玻璃,基于所提方法匹配线结构光测量的玻璃数据与三坐标数据,两者偏差在-0.06/0.08 mm,基本满足工业需求。Objective The registration problem of continuous surfaces with weak features poses a challenge in computer vision and image processing. Automotive glass is a typical example of such a continuous surface with weak features. Due to the unclear characteristics of the Reference Point System(RPS) on automotive glass, it is difficult to accurately register the three-dimensional data obtained from either three-coordinate measurement or optical methods to the RPS coordinate system. To address this issue, the research proposes a low-cost and progressive registration algorithm that does not rely on high-precision fixtures and can still achieve precise registration and dimensional evaluation.Methods This method first builds upon the "point to surface" ICP registration, and further proposes a rough matching with boundary penalty function correction to achieve initial alignment between the measurement data and the model, providing good initial values for subsequent registration(Fig.1). Secondly, in order to align with the results of CMM's measurement and meet industrial needs, the distance between non-RPS sampling model points and corresponding measurement data points is directly optimized through CMM's evaluation method, and the measurement point cloud is fine tuned. In order to adjust the overall minimum registration benchmark from the previous step to the RPS benchmark(Fig.2), the distance between the RPS point and the corresponding point in the measurement data was directly optimized using CMM's evaluation method, thereby achieving accurate positioning and surface shape evaluation of the RPS point in the measurement data of weak feature continuous surfaces.Results and Discussions This article validates the effectiveness of the proposed progressive RPS point positioning algorithm through simulation and actual experiments. Combined with(Fig.11), it can be clearly seen that each step of adjusting the measurement data makes the deviation between the measurement data and the model closer to the deviation between the three coor

关 键 词：仪器科学与技术 弱特征表面配准 渐进式匹配算法 汽车玻璃 ICP算法 

分 类 号：TH741[机械工程—光学工程]