MRI: Acquisition of Light Detection and Ranging (LIDAR) Scanner for Computer Vision Research and Interdisciplinary Education

该项目获得了一台徕卡ScanStation P30，旨在开发对齐二维（2D）和三维（3D）数据的方法，并使用生成的3D可视化来组装教育工具。激光雷达（光探测和测距）扫描仪将被在该机构学习和从事计算机科学、电影和艺术领域研究的教授和学生使用。该小组以研究人员现有的工作为基础，通过3D激光雷达扫描注册的2D图片和视频集创建可导航的3D环境。有了这些工具，这个项目将实现以下相互交织的目标。(i)建立跨学科研究小组，进一步研究如何描述和登记多模态数据，或用不同相机和传感器获得的数据。（ii）利用已开发的方法与电影和历史方面的合作者建立教育3D应用程序。开发新的跨学科课程，利用拟议的扫描仪使许多学生有机会接触到最先进的技术。开发的课程将帮助调查员为她的研究实验室培训和招募学生，这些学生可以从事跨学科项目。虽然目前收集和存储2D图像和3D激光雷达扫描可能更容易、更经济，但融合这种多模态数据仍然具有挑战性。由于在人工场景中经常出现的重复和模糊，以及同一主题不同渲染上的各种属性，在自动匹配方法中经常产生混淆。在一段时间内收集的图像集，在此期间照明条件和场景内容可能发生变化，不同的艺术渲染以及不同的传感器类型（相机，激光雷达等），都可能导致数据集的视觉变化。然而，通过结合上下文信息来可视化每个图像源中直观存在的区域属性，支持者已经成功地解决了许多障碍。她的研究小组还将研究如何将人类感知用于将3D激光雷达数据与其他类型的视觉信息进行注册。

This project, acquiring a Leica ScanStation P30, aims to develop methods for aligning two dimensional (2D) and three dimensional (3D) data and to assemble educational tools using the produced 3D visualizations. The LIDAR (LIght Detection And Ranging) scanner will be used by professors and students studying and performing research in the fields of computer science, film, and art at the institution. The group builds on the researcher's existing work that creates navigable 3D environments from sets of 2D pictures and videos that are registered with 3D LIDAR scans. With the instrumentation this project would enable the following intertwined goals.(i) Build an interdisciplinary research group to further study how to describe and register multi-modality data, or data obtained with different cameras and sensors.(ii) Use the developed methods to build educational 3D applications with collaborators in film and history.(iii) Develop new, cross-disciplinary courses that use the proposed scanner to provide access for many students to the state-of-the-art technology.The developed courses would help the investigator train and recruit students for her research lab that can work on interdisciplinary projects.Although at present collecting and storing both 2D images and 3D LIDAR scans might be easier and more economical, fusing such multi-modality data remains challenging. Confusion is often created in automatic matching methods by the repetition and ambiguity that often occur in man-made scenes as well as the variety of properties on different renderings of the same subject. Image sets collected over a period of time during which lighting conditions and scene content may have changed, different artistic rendering, as well as varying sensor types (Camera, LIDAR, etc.), can all contribute to visual variations in data sets. However, by incorporating contextual information to visualize the regional properties that intuitively exist in each imagery source, the proponent has successfully addressed many of the obstacles. Her research group will also study how human perception can be used for registering 3D LIDAR data with other types of visual information.