CAREER/CDS&E: Advanced, 3D Infrastructure Information Modeling Using Lidar

职业/CDS

• 批准号: 1351487
• 负责人: Michael Olsen
• 金额: $ 40万
• 依托单位: Oregon State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-04-01 至 2020-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1351487&HistoricalAwards=false
• 关键词: CAREER; CDS& Advanced; 3D; Infrastructure

The primary research focus of this Faculty Early Career Development (CAREER) Program award is to efficiently identify and extract meaningful information from three-dimensional, geospatial models of transportation infrastructure in a holistic, automated framework, enabling broader application. Advanced mapping technologies such as laser scanning produce three-dimensional maps, creating highly detailed scenes that can be virtually explored and queried for a diverse range of purposes including infrastructure management, digital terrain modeling, cultural heritage, flood plain delineation, and landslide detection. However, tradeoffs exist between the detail and scale provided by these technologies and the immense size of the resulting datasets. This complexity can strain the most powerful computational resources and require a steep learning curve to exploit the data. While recent tools have made significant progress, only a small and piece-meal portion of information can be automatically extracted from these rich datasets compared to what is actually available. Key scientific questions to be addressed through this research include (1) What inherent attributes of an object and associated representation in laser scan data and supporting imagery are most beneficial to accurately identifying and extracting an object?, (2) How can neighboring features and context of an object help with rapidly identifying it within geospatial data?, and (3) How can an abridged framework be developed to improve information extraction from laser scan data to consider the broad range of transportation objects? This overarching framework will consider a broad range of object types, incorporate advanced system information and data structuring, function in noisy, real-world environments, and focus on datasets covering large spatial scales consistent with transportation infrastructure management. Products resulting from this framework include a transportation infrastructure object properties database, fully-classified benchmark datasets, new algorithms, and supporting code, which will be made publicly available. Well-maintained transportation infrastructure is vital to our economy as well as public safety. Most transportation agencies charged with maintaining infrastructure are trying to develop a comprehensive methodology for inventory, maintenance and management of their immense assets. In many cases, the available resources are reduced while maintenance demands still increase. This research will provide timely solutions to map and digitally manage these assets more efficiently and cost-effectively than current practices. Although primarily focused on transportation, the computational methods and techniques will be applicable and extendable to a wide range of other applications such as land management, urban mapping, and robotics. This project also will provide students with multi-disciplinary education and training in geospatial analysis, computer science, transportation, and engineering. Despite the high demand for geospatial expertise today, educational opportunities are limited and challenging because of the rapid evolution of the supporting technologies. As a result, the U.S. has an insufficient number of geospatially-trained students entering the workforce to meet the ever-increasing demand utilizing geospatial information throughout society. This project will enhance geospatial education through activities ranging from exposure at public events to training camps for high school age students to creation of a model civil engineering geomatics graduate program.

这项教师早期职业发展（Career）计划的主要研究重点是在一个整体的自动化框架中，有效地从交通基础设施的三维地理空间模型中识别和提取有意义的信息，从而实现更广泛的应用。先进的测绘技术，如激光扫描产生三维地图，创建高度详细的场景，可以虚拟探索和查询各种目的，包括基础设施管理，数字地形建模，文化遗产，洪泛平原划定和滑坡检测。然而，在这些技术提供的细节和规模与最终数据集的巨大规模之间存在权衡。这种复杂性可能会耗尽最强大的计算资源，并且需要一个陡峭的学习曲线来利用数据。虽然最近的工具取得了重大进展，但与实际可用的信息相比，只有一小部分信息可以从这些丰富的数据集中自动提取出来。通过本研究要解决的关键科学问题包括：(1)物体的哪些固有属性及其在激光扫描数据和支持图像中的相关表示最有利于准确识别和提取物体？(2)在地理空间数据中，物体的相邻特征和上下文如何帮助快速识别该物体？(3)如何开发一个简化的框架来改进激光扫描数据的信息提取，以考虑广泛的运输对象？这个总体框架将考虑广泛的对象类型，结合先进的系统信息和数据结构，在嘈杂的现实环境中发挥作用，并专注于覆盖与交通基础设施管理一致的大空间尺度的数据集。该框架产生的产品包括交通基础设施对象属性数据库、完全分类的基准数据集、新算法和支持代码，这些都将公开提供。维护良好的交通基础设施对我们的经济和公共安全至关重要。大多数负责维护基础设施的运输机构都在努力开发一种全面的方法来清点、维护和管理其庞大的资产。在许多情况下，可用资源减少，而维护需求仍然增加。这项研究将提供及时的解决方案，比目前的做法更有效、更经济地绘制和数字化管理这些资产。虽然主要集中在交通运输，计算方法和技术将适用和扩展到广泛的其他应用，如土地管理，城市测绘和机器人。该计划也将为学生提供地理空间分析、计算机科学、交通运输和工程等多学科的教育和培训。尽管今天对地理空间专业知识的需求很高，但由于支持技术的快速发展，教育机会有限且具有挑战性。因此，美国没有足够的地理空间训练的学生进入劳动力市场，以满足整个社会对地理空间信息日益增长的需求。该项目将通过各种活动加强地理空间教育，包括在公共活动中为高中学生提供培训，以及创建一个土木工程地理信息研究生课程的模型。