提高结构光三维测量几何精度与分辨率的方法研究

Geometric accuracy and resolution have always been the key technical qualifications and research focus of the structured-light (SL) three-dimensional measurement system. In the SL system, geometric data is generated by a three-step process—projecting coded fringes, sensor imaging and stereo vision reconstruction. The current methods to improve the geometric accuracy usually try to decrease single-stage errors without considering cross-stage errors, while the current methods to improve the geometric resolution usually exploit complex methods or devices such as data fusion with sensors of different principles and physical fringes. In this project, we will research on a cross-stage error model of the SL 3D measurement system based on Bayesian statistics and make strategy to compensate errors and improve the geometric accuracy. We will design a multi-projector and multi-camera SL 3D measurement device and use geometric super-resolution method to reconstruct high resolution geometric data from multiple measurements so as to improve geometric resolution under a fixed imaging distance. We will design an experiment using very large scale objects and very small objects as targets to testify the uncertainty relationship between geometric accuracy and resolution under extreme situations. The results of the research project provide theoretical support for designing the geometric accuracy and resolution specifications of a SL system as well as provide effective technical methods to improve the geometric accuracy and resolution performance of existing 3D measurement devices.

几何精度与分辨率始终是结构光三维测量系统的关键技术指标与研究焦点。在结构光系统中，由投影机投射编码光栅，历经相机成像、立体视觉重建生成几何数据；现有提高几何精度的研究大多针对某种具体的投影引入误差或相机引入误差进行单独补偿，未考虑不同类型误差耦合对几何精度的影响；现有提高几何分辨率的研究大多采用异构传感器数据融合或物理光栅等复杂手段或设备。本项目研究内容包括：建立基于贝叶斯统计理论的结构光三维测量的耦合误差模型，据此制定误差补偿方案，提高几何精度；设计多投影机多相机的结构光三维测量装置，采用几何超分辨率方法从多次测量数据中重建高分辨率几何数据，提高固定观测距离下的几何分辨率；以大尺寸物体及微小物体为实验测量对象，验证极限情况下几何精度与分辨率的测不准关系。本项目研究成果为设计结构光系统的几何精度与分辨率指标提供针对性的理论支持，为提高现有三维测量设备的几何精度与分辨率提供一种有效技术手段。

几何精度与分辨率始终是结构光三维测量系统的关键技术指标与研究焦点。在结构光系统中，由投影机投射编码光栅，历经相机成像、立体视觉重建生成几何数据；现有提高几何精度的研究大多针对某种具体的投影引入误差或相机引入误差进行单独补偿，未考虑不同类型误差耦合对几何精度的影响；现有提高几何分辨率的研究大多采用异构传感器数据融合或物理光栅等复杂手段或设备。本项目研究内容包括：建立基于贝叶斯统计理论的结构光三维测量的耦合误差模型，据此制定误差补偿方案，提高几何精度；设计多投影机多相机的结构光三维测量装置，采用几何超分辨率方法从多次测量数据中重建高分辨率几何数据，提高固定观测距离下的几何分辨率；以大尺寸物体及微小物体为实验测量对象，验证极限情况下几何精度与分辨率的测不准关系。本项目研究成果为设计结构光系统的几何精度与分辨率指标提供针对性的理论支持，为提高现有三维测量设备的几何精度与分辨率提供一种有效技术手段。