Automatic Recognition and Location of Civil Infrastructure Objects from Image Sequences

从图像序列中自动识别和定位民用基础设施对象

• 批准号: 9812783
• 负责人: Rongxing (Ron) Li
• 金额: --
• 依托单位: Ohio State University Research Foundation -DO NOT USE
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-09-01 至 2001-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9812783&HistoricalAwards=false
• 关键词: Automatic; Recognition; Location; Civil; Infrastructure

9812783Developments in information technology are necessary to support civil infrastructure systems that are large, complex, distributed, dynamic systems including communications, roads, bridges, and distribution of water, gas and electricity. One of the most important data sources for such systems is updated spatial locations, physical conditions, and other attributes. These data describe dynamic status changes of objects caused by, for example, usage and deterioration of materials, repair and relocation, climate conditions, accidents, natural hazards, and other unexpected events. The new technology of mobile mapping systems integrates GPS receivers, INS (inertial navigation system) and stereo CCD (coupled charges device) cameras on a mobile platform for rapid high quality spatial data acquisition. Infrastructure objects appearing in a georeferenced mobile mapping images can be measured on a computer screen and their 3-D ground locations are calculated from measured 2-D image coordinates using a photogrammetric model. Such a technology is extremely efficient, but it demands a high level of automation and intelligence in data processing for object recognition and location. In this research: 1) a neural network model will be developed to simulate an operator's vision and measurement capability and to automate the procedure of recognition and location of objects; 2) the relationship between known infrastructure objects, their image features, and neural network characteristics will be studied for developing networks capable of recognizing complex objects; and 3) spatial civil infrastructure information databases will be generated automatically from the image sequences. Knowledge algorithms will be developed first for transportation systems, and then will be extended to other civil infrastructure objects.This research will result in significant progress in spatial information generation to support information technology-based civil infrastructure planning, design, construction, maintenance and repair and management. It will demonstrate an innovative integration of space technology and image processing for civil infrastructure applications. The spatial information databases generated can also be input to a GIS where infrastructure information can be integrated with other information, such as transportation networks, geological models, environmental layers, and others to support complex interdisciplinary decision making processes.****

9812783信息技术的发展对于支持民用基础设施系统是必要的，这些系统是大型的、复杂的、分布式的、动态的系统，包括通信、道路、桥梁以及水、气和电的分配。 这些系统最重要的数据源之一是更新的空间位置，物理条件和其他属性。 这些数据描述了由例如材料的使用和劣化、修理和重新安置、气候条件、事故、自然灾害和其他意外事件引起的对象的动态状态变化。 移动的测绘系统的新技术将GPS接收机、INS（惯性导航系统）和立体CCD（耦合电荷器件）相机集成在一个移动的平台上，以快速获取高质量的空间数据。 可以在计算机屏幕上测量出现在地理参考的移动的测绘图像中的基础设施对象，并且使用摄影测量模型从测量的2-D图像坐标计算它们的3-D地面位置。 这种技术非常有效，但它需要在对象识别和定位的数据处理中实现高度自动化和智能化。 在这项研究中：（一）发展一个神经网络模型，以模拟操作员的视觉和测量能力，并使物体的识别和定位过程自动化;（二）研究已知基础设施物体、其图像特征和神经网络特征之间的关系，以发展能够识别复杂物体的网络;（3）从图像序列中自动生成空间民用基础设施信息数据库。 知识算法将首先针对交通系统开发，然后将扩展到其他民用基础设施对象，这项研究将导致空间信息生成的重大进展，以支持基于信息技术的民用基础设施规划，设计，施工，维护和维修和管理。 它将展示空间技术和图像处理在民用基础设施应用方面的创新性结合。生成的空间信息数据库也可以输入到GIS中，在GIS中，基础设施信息可以与其他信息（如交通网络，地质模型，环境层等）集成，以支持复杂的跨学科决策过程。