CAREER: Visual Pattern Recognition Models for Remote Sensing of Civil Infrastructure

职业：民用基础设施遥感视觉模式识别模型

• 批准号: 0948415
• 负责人: Reginald DesRoches
• 金额: $ 40.23万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2010
• 资助国家: 美国
• 起止时间: 2010-09-01 至 2013-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0948415&HistoricalAwards=false
• 关键词: CAREER; Visual; Pattern; Recognition; Models

The research objective of this Faculty Early Career Development (CAREER) award is the creation of pattern recognition models that will automate the recognition of civil infrastructure-related elements based on the knowledge of 3D surfaces from remote sensing, and visual features from pattern recognition. In other words, the intellectual merit of this project lies in the creation of the missing link between remote sensing and pattern recognition that will automate the transformation of 3D surfaces into information rich, 3D element models with the help of machine vision. The main contribution of this project is that, instead of manually recognizing each element every time it is encountered, we need only recognize its characteristics once and automatically detect it each subsequent time. This is analogous to defining an alphabet (letters = characteristics) so that this project will build the words (element models) and find them in a text (3D surface), instead of having to manually find the words in every text we encounter. The benefit comes from the ability to reuse the known letters (characteristics) and words (element models) every time we have a new text (3D surface).The immediate advantage that will result from this work is the ability to automate the element recognition step of the "as-built" model generation process. The National Academy of Engineering recently listed "Restoring and Improving Urban Infrastructure" as one of the Grand Challenges of Engineering in the 21st century. Two of the greatest issues that cause this grand challenge are the need for more automation in construction, through advances in computer science and robotics, and the lack of viable methods to map and label existing infrastructure. Over two thirds of the effort needed to model even simple infrastructure is spent on manually converting surface data to a 3D model. The result is that as-built models are not produced for the vast majority of new construction and retrofit projects, which leads to rework and design changes that cost up to 10% of the installed costs. Any efforts towards automating the modeling process will increase the percentage of infrastructure projects being modeled and, considering that construction is a $900 billion industry, each 1% of increase can lead up to $900 million in savings.

该学院早期职业发展奖的研究目标是创建模式识别模型，该模型将基于来自遥感的3D表面知识和来自模式识别的视觉特征自动识别与民用基础设施相关的元素。换句话说，这个项目的智力价值在于创造了遥感和模式识别之间缺失的一环，这将在机器视觉的帮助下自动将3D表面转换为信息丰富的3D元素模型。这个项目的主要贡献是，我们只需要识别一次它的特征，然后每次随后自动检测它，而不是每次遇到它时都手动识别它。这类似于定义字母表(字母=特征)，以便此项目将构建单词(元素模型)并在文本(3D表面)中找到它们，而不是必须在我们遇到的每个文本中手动查找单词。好处来自每次我们有新文本(3D表面)时重用已知字母(特征)和单词(元素模型)的能力。这项工作将产生的直接好处是能够自动执行“建成”模型生成过程的元素识别步骤。美国国家工程院最近将“恢复和改善城市基础设施”列为21世纪工程学的重大挑战之一。造成这一重大挑战的两个最大问题是，通过计算机科学和机器人技术的进步，需要在建筑中实现更多的自动化，以及缺乏可行的方法来绘制和标记现有的基础设施。为简单的基础设施建模所需的工作中，超过三分之二的工作都花在了手动将表面数据转换为3D模型上。其结果是，绝大多数新建和改造项目没有生产竣工模型，这导致返工和设计更改，成本高达安装成本的10%。任何自动化建模过程的努力都将增加正在建模的基础设施项目的百分比，考虑到建筑业是一个价值9000亿美元的行业，每增加1%，就可以节省高达9亿美元的成本。