1 Extracting Surveying Goals from Point Clouds to Support Construction and Infrastructure Inspection Pingbo Tang 1 , and Burcu Akinci 1 1 Civil and Environmental Engineering Department, Carnegie Mellon University Abstract Inspectors need accurate and efficient methods for obtaining geometric information about facilities for construction and infrastructure management. Laser scanners can capture millions of 3D points, which have mm accuracy, in minutes, and hence can streamline extraction of geometric information. Currently, extraction of geometric information is done manually. This paper describes a case study, which shows that manual approach is inefficient and error-prone. Aiming at improving the efficiency and effectiveness of extracting survey goals, this paper discusses computer interpretable representations of surveying goals as a first step towards extraction of surveying goals. The described approach targets enabling surveyors to compose surveying goals as <facility parts, features, operations> tuples. Reasoning mechanisms can transform user-defined surveying goals described by these tuples into a sequence of measurement operations to generate results. Keywords: Laser Scanning; Information Retrieval; Construction and Infrastructure Inspection, Surveying Goal Introduction Construction and infrastructure inspectors need accurate and efficient methods for obtaining geometric information about facilities. Laser scanners can collect millions of 3D points in minutes with mm-level accuracy in the form of dense point clouds (Zoller+Fröhlich 2005), and hence provide data to extract detailed geometric information. At the same time, laser scanned data also brings unique challenges for geometric information retrieval. State-of-art 3D reverse engineering environments, such as Polyworks (InnovMetric 2005), enable inspectors to manually conduct measurement operations, such as extraction of geometric features (e.g. plane, edge and point) and spatial relationships between geometric features (e.g. distances between points). However, a case study described in this paper indicates that many inspection tasks, such as obtaining the minimum vertical underclearance of a bridge, require inspectors to execute a few types of measurement operations repetitively. This not only reduces the efficiency of the information retrieval process, but also causes subjectivity and reliability issues due to being a manual process. The research described in this paper targets streamlining extraction of surveying goals (e.g., the minimum vertical underclearance of a bridge) by enabling capture and representation of surveying goals in a computer interpretable way. With formal representations developed in this research, inspectors can compose surveying goals as <facility parts, features, operations> tuples. Using several geometric reasoning mechanisms that are currently being developed, it is possible to reason with surveying goal tuples to generate and automatically execute operation sequences. This results in