Pingbo Tang's Home Page

PhD Candidate
Advanced Infrastructure System Group
Civil and Environmental Engineering Department
Carnegie Mellon University
Tel: 1+412-268-6271
Email: tangpingbo@gmail.com

Notice: I have moved my personal web page to http://sites.google.com/site/tangpingbo/ , this page will automatically transfer you to the new page within 10 seconds.

Geometric Data Collection and Automated Data Interpretation for Construction and Infrastructure Management

Hello! My name is Pingbo Tang. Tang is one of the greatest dynasty in Chinese history, Ping means calm, gentle, flat, and mild in Chinese, and Bo means wavy and dynamic in Chinese.  Of course, I am made in China :-), my hometown is Hengyang, a city in Hunan. In 2005, after I finished my master thesis about Bridge CAD at Tongji University, China, P.R., I became a PhD research assistant of Professor Burcu Akinci's MOSAIC (Management of mOdel-based Sensor-driven Advanced Infrastructure and Construction Systems) group at Carnegie Mellon University.  MOSAIC is part of Advanced Infrastructure System Group here at Carnegie Mellon. Here are my resume and CV.

Basically I am interested in Bridge CAD/CAE (Computer-Aided-Design/Computer-Aided-Engineering) and Bridge Management. What I am trying to do is to apply various information and communication technologies (ICT), such as remote sensing, computer vision, knowledge-based system to enhance people's capability of assessing the condition of buildings and infrastructure systems and plan preventive maintenance activities accordingly. Currently I am focusing on automated 3D data interpretation for construction and infrastructure management.I am tesing the technical feasibility of using laser scanners to collect geometric data of building structures and bridges to support construction and infrastructure inspection. Laser scanning scanners are 3D imaging cameras which can collect highly accurate and dense 3D point clouds of a building or infrastructure in minutes. Combined with computer vision and semantically rich Building Information Models (BIM),  I am trying to automatically recognize objects such as beams, columns from the 3D point clouds, and developing knowledge-based systems to support automated 3D data interpretation so that inspectors can efficiently extract spatial information efficiently from huge amount of points to answer questions such as "which columns are not plumb?", "which part of the slab surfaces do not satisfy the flatness requirement?", "Are all walls at the correct position?" etc. Augmenting large 3D data sets with semantic information and knolwegde-based reasoning mechanisms, this research will be of great potential for the automation of construction and infrastructure inspection process.

One important aspect of my research interest is automatic reconstruction of as-built semantically rich CAD model from point clouds, it somehow serve as  a foundation for automatic data interpretation.
Point clouds without semantic and semantically rich bridge model with component information, topological relationship (connections etc.) and so on. If we can automatically transform points to such semantically rich Building information model, you can imagine how people can deal with millions of points easily. They finally need to know which component is OK, which is not, and do not want to check millions of points without any intelligence, right?
      

Another important aspect of my research is accuracy analysis of laser scanned data, rather than point-accuracies, I am more interested in and how point accuracies influence the accuracies of features extracted from point clouds (such as edges, parametric surfaces), how the accuracies of measurements (angles, distances) relying on those features (such as orientation of a plane) are influenceed by point accuracy and feature accuracy. Those questions are directly related to how accurately inspectors can expect some measurements can be extractected from point clouds, and how they can actively get most out of the data by properly dealing with some negative effects on measurement accuracies (such as noisy data).