A preliminary investigation into automated identification of structural steel without a priori knowledge

Abstract

One of the most prohibiting factors when attempting to reuse structural steel members or systems of members is the time and labour required to accurately determine dimensions. Current practices dictate that all required measurements are recorded by hand using tape measures and calipers which adds a significant cost to reused steel. To mitigate this cost, a semi-automated method for identifying structural steel components and systems is proposed that uses data acquired in the form of a 3D point cloud. Current research in the field of automated object recognition currently has two major limitations: (1) a priori knowledge, such as a building information model (BIM) is required, or (2) only simple, flat surfaces can be identified. The purpose of this study is to preliminarily investigate the possibility of automating the process of (1) cross section identification, (2) end connection geometry of bolted connections, and (3) relative component position of multi-component, planar structural systems such as trusses. Cross section identification is performed by creating filters that match standard structural sections and then convolving them over images of the cross section data. The end connection geometry is identified using Hough algorithms to detect lines and circles representing the limits of the component and the bolt holes, respectively. Planar structural systems are identified using Hough algorithms to detect lines which represent the components of the system. The results from the proposed methods show a strong potential for fully automated processes to be able to identify structural steel components and systems without a priori knowledge.