强反射表面三维扫描测量关键技术研究

The strong reflection surface belongs to the non-Lambertian surfaces, which easily makes the over-exposure of the digital camera, and resulting in the loss of measurement information. Meanwhile, the limitation of speed and time resolution affects the realization of dynamic three-dimensional shape measurement. More importantly, differential visual measurement method has not been proposed by far, and the anti-interference ability of three-dimensional scanning measurement system is generally poor. In view of the above problems, the main research contents of this project are as follows: firstly, three-dimensional geometric parameter extraction method based on high dynamic range imaging in the local highlight background is proposed, including scattering characteristics and mechanism of non-Lambertian surfaces, response model and calibration of imaging system, adaptive high dynamic range imaging and tone mapping. Secondly, dynamic three-dimensional shape measurement method based on time-domain pixel multiplexing is put forward, including modulation and coding of time information of incident light, fast extraction and interpolation of pixels, high-speed and high-resolution imaging and dynamic shape measurement. Then the differential three-dimensional geometric feature measurement technology based on dual projection is also presented, including differential measurement structure, adaptive calibration, and the coordination strategy and measurement consistency of dual projection differential measurement. Finally, construction and experimental research of measurement system and its application are described in detail, including computational imaging system, structural optimization, the overall experimental design and validation, and application adaptability. In short, the project can solve the key problems of three-dimensional scanning measurement of strong reflection surface, including 3D point cloud loss caused by partial over-exposure, dynamic three-dimensional shape measurement, and the anti-interference ability of three-dimensional scanning measurement system. The project will provide new detection means for manufacturing field.

强反射表面属于非朗伯表面，扫描测量时易使相机曝光量饱和，造成测量信息缺失；速度与时间分辨率的限制，影响了动态三维面形测量；尚未见差动式视觉测量方法，测量抗干扰能力差。针对上述问题，本项目主要研究内容有：①基于高动态范围成像的局部高光背景中三维几何参数提取方法：非朗伯体表面散射特性和作用机理、成像系统响应模型与标定、自适应高动态范围成像和色调映射。②基于时域像素复用的动态三维面形测量方法：入射光线时间信息的调制与编码、像素快速提取与插值组合、高速高分辨率成像和动态面形测量。③基于双投影差动式三维几何特征测量技术：差动测量结构、自适应标定、双投影差动协同策略与测量一致性。④测量系统构建与实验研究及应用探索：计算成像系统、结构优化配置、整体实验设计与验证、应用适用性。项目解决局部过曝光造成的三维点云缺失、动态三维面形测量、抗干扰能力等强反射表面三维扫描测量关键问题，为制造领域提供新的检测手段。

项目的背景：强反射表面三维扫描测量技术具有很高的工业应用价值，但非朗伯表面的局部强反射会在工业相机中形成高亮区域，使得表面信息缺失；对于动态物体或场景的测量通常需要高速设备、大带宽和存储空间；在结构光三维扫描测量中，会面临一些复杂场景，噪声会干扰数据的采集。本项目就针对以上问题进行研究。.主要研究内容：强光表面高动态范围成像及工件检测、强光物体三维测量、动态三维测量和差动式三维测量技术。.重要结果和关键数据：①在高动态范围成像研究方面，搭建基于数字微镜器件（Digital Micromirror Device, DMD）的计算成像系统，进行系统的匹配、映射与标定，实现自适应高动态范围（High Dynamic Range, HDR）成像，制作了金属表面缺陷数据集，引入深度学习，实现了样品在非强光下的缺陷识别率为99.5%，强光下的缺陷识别率为85.0%。使用频域复用原理，将不同曝光度图像置于不同子频带处，用带通滤波将其分离，而后将分离后图像融合恢复成一幅HDR图像，融合图像与原HDR图像的PSNR达31.12dB。②将HDR计算成像方法用于结构光高亮物体三维测量，提出了强光补偿方法，恢复被高光淹没的相位信息，对高亮的标准球∅40.056mm和金属工件30.056mm×28.332mm进行三维测量实验，球直径、金属工件长和宽的测量误差分别为0.080mm、-0.103mm和-0.094mm。③在视觉动态测量研究方面，进行了计算成像系统的时分复用方法研究，使用数字投影仪对被测物体同时进行时间和空间编码，使一幅图像记录下多个时刻序列信息；而后，从图像中分离出不同时刻的编码信息，实现用低速工业相机记录高速运动，频率测量达到200Hz、误差小于1Hz。④搭建了双投影单相机系统，提出相位互补、六图分离和彩色条纹分离等算法，实现叠加光栅的分离，提高了测量效率，增强了系统的抗干扰能力。.科学意义：项目为强反射表面三维扫描测量中若干关键问题提出了解决方案，如工业中强反射表面高亮区域信息缺失、高亮工件缺陷检测、动态三维测量速度慢等，为制造领域提供新的检测手段。