雪雕艺术品的3D视觉测量关键技术研究

This research studies the measuring technology of structured light in outdoor environment, on the basis of the digital 3D reconstruction of snow sculptures. Without changing the structure of structure light sensors, this research aims at enhancing the anti-noise ability of the system by means of digital image processing, and ensuring the accuracy and speed of the experimental results, so that the sensors can be applied in outdoor environment. As for the reflective property and ambient illumination noise of the surface of snow sculptures, the applicant plans to design the RGB weighting function according to the noise classification results in the process of the preprocessing of images, and transforms the color image into the monochrome image to eliminate most of the ambient noises. And then based on the spatial-temporal context information in the video sequence, the applicant constructs a spatial-temporal Hidden Markov model with the adaptive parameter to extract and track the light strip of structured light. The model is insensitive to the obstinate noise and reflected light of the image, so the extracted light strip is accurate, continuous and smooth. After transformed through the coordinate system, the tracking path of the light strip can be taken as the 3D measuring point cloud of snow sculptures. In the process, the key questions (noise interference, light strip deformation and other questions) of structured light are concentrated on a spatial-temporal tracking framework and the measuring results are of robustness, accuracy and fast.

本研究以雪雕艺术品的数字化3D重建为背景，研究野外环境下的结构光测量技术。旨在不改变结构光传感器结构的前提下，通过数字图像处理手段提高系统抗噪能力，保证测量结果的精度和实时性，使传感器能够作业于野外复杂环境中。针对雪雕表面的反光特性和环境光照噪声，申请人拟在图像预处理过程中，通过噪声分类结果设计RGB颜色权重函数，将彩色图像转换为单色值图像，以消除大部分环境噪声。然后，基于视频序列中的时空上下文信息，构建一套自适应参数的时空隐马尔可夫模型，提取和跟踪结构光光条。模型对图像中的顽固噪声和反光不敏感，提取的光条结果具有精确、连续、光滑的特点；光条跟踪轨迹经坐标系转换即可作为雪雕的3D测量点云。整个处理过程将结构光图像中的关键问题（噪声干扰、光条形变等）集中到一个时空跟踪框架中，测量结果兼顾鲁棒性、准确性和实时性。

项目以3D视觉测量雪雕艺术品为目标，研究野外环境下的结构光测量技术。旨在不改变结构光传感器结构的前提下，通过数字图像处理手段提高系统抗噪能力，保证测量结果的精度和实时性，使传感器能够作业于野外复杂环境中。针对雪雕表面的反光特性和环境光照噪声，申请人拟在图像预处理过程中，通过噪声分类结果设计RGB颜色权重函数，将彩色图像转换为单色值图像，以消除大部分环境噪声。然后，基于视频序列中的时空上下文信息，构建一套自适应参数的时空隐马尔可夫模型，提取和跟踪结构光光条。模型对图像中的顽固噪声和反光不敏感，提取的光条结果具有精确、连续、光滑的特点；光条跟踪轨迹经坐标系转换即可作为雪雕的3D测量点云。整个处理过程将结构光图像中的关键问题（噪声干扰、光条形变等）集中到一个时空跟踪框架中，测量结果兼顾鲁棒性、准确性和实时性。研究成果在直线、平行线和网格模式的结构光传感器上验证，进而为不同应用平台（如固定式或手持式结构光扫描仪）的三维扫描、三维重建等提供3D测量的理论基础和技术支持。项目解决了未知环境下图像多源噪声解析，时空模型的自适应和联合优化求解等科学问题。