Automated rebar diameter classification using point cloud data based machine learning

Abstract

Inspecting the diameter and spacing of rebar is an important task conducted by fabricators and site engineers during the manufacturing and construction stages. This is because the bearing capacity of reinforced concrete structures is affected by the size and position of the rebar, so installing rebar of the correct size and position should be ensured to safeguard the structural integrity of the structure. This study presents a new terrestrial laser scanning (TLS)-based method using machine learning to automatically classify rebar diameters and accurately estimate rebar spacing. To this end, a new methodology, named density based machine model, is proposed to improve classification accuracy. To validate the proposed method, experimental tests on laboratory specimens with rebars of seven different diameters are conducted. The results show that the prediction accuracy for large rebar diameters measuring D25-D40 are up to 97.2%, demonstrating great potential for the application of the proposed technique on manufacturing and construction sites. The key findings of the study are: (1) the proposed density-based modeling method for rebar diameter prediction is superior to the traditional machine learning approach; (2) scan density is one of the most important factors in the prediction results, especially in the small rebar diameter group; and (3) it was found that at least 10 points/cm2 is necessary to ensure accurate rebar diameter classification in small rebar diameters between D10-D20. It is expected that the proposed rebar diameter and rebar spacing technique will be useful in providing autonomous and accurate rebar inspection in manufacturing factories and on construction sites.