Semiautomated detection of design errors in 2D drawings using 3D reconstruction

This article reports an extension of a previous study on developing a systematic approach to construction rework diagnosis and reduction (Li and Love, 1998; Love, Li, and Mandal, 1998a,b; Love, Smith, and Li, 1998). Existing studies identified that design errors are a primary source (cause) of rework in construction, contributing to 51 % of the total failure costs of rework (Li and Love, 1998; Hammarlund and Josephson, 1991). Among these design errors, most of them are attributable to dimensional inconsistencies presented in different sets or sections of engineering drawings (Mondal et al., 2000). In this article, we present a set of virtual reality (VR) enabled 3D reconstruction and collision-detection techniques to automatically identify and minimize design errors. The 3D reconstruction technique can reconstruct 3D models from three 2D orthographic views. This process helps identify many dimensional inconsistencies within different orthographic views of the same object that give rise to incomplete/irregular 3D models. The 3D reconstruction technique is based on the algorithm developed in Wang et al. (2001) for automatic reconstruction of 3D models from three 2D orthographic views. The 3D models provide input to the collision detection study for identifying inter-object dimensional inconsistencies. Thus, the integrated application of 3D reconstruction and the collision-detection techniques is able to identify all the dimensional inconsistencies within and between objects. In this article, the 3D reconstruction technique and the collision detection technique are described through their applications in handling construction drawings.