Shape Recognition with Point Clouds in Rebars

Abstract

Purpose In this paper, the authors describe the methods of inspecting the quality of reinforced concrete structure using point-clouds data acquired from a 3D-laser scanner. A 3D-laser scanner is an outstanding device to analyze a realworld object and to collect digital data on its shape. Inspections of the quality of reinforced concrete structure using point clouds are required for novel methods which count the quantity of rebar material and to check the space of each rebar.Method To inspect with the use of 3Dpoint clouds, we developed a method of 3D-point-clouds recognition of the rebars elements. In this paper the authors show three methods to analyze point clouds collected in rebars of building structures. Firstly, the authors developed a method of noise reduction to make a clear distinction between object surface points and noise points; this is important because point clouds in rebars have much noise that can disturb subsequent analysis. Secondly, the authors developed a method of abstracting point clouds on reinforcement bars. Thirdly, the authors developed a method dividing point clouds on reinforcement bars into hoops and wall horizontal reinforcement bars. The authors scanned reinforced bars of columns and walls at a construction site of an apartment and then applied the three methods to analyze the point clouds data. Results & discussion In this experiment, our methods were able to identify 3D point clouds as main bars, horizontal bars, and hoops. We were able to measure the object from 3D-point-clouds data at any time as well as being able to develop an automatic inspection system.

Tracking of secondary and temporary objects in structural concrete work

“…Furthermore, their work did not address the issue of associating the recognition of each formwork (temporary object) to a particular primary design object, which is critical to estimate progress accurately, nor did their work address scaffolding or rebar. The second relevant research is the work by Ishida et al (2012) who developed a system to inspect CI 14,2 the quality of a concrete reinforcement using 3D point clouds obtained with TLS. Essentially, they used a shape recognition technique to detect steel reinforcing bars (rebar) (this is preceded by the application of a noise filter to the 3D point cloud).…”

“…The integration of colour information within the recognition framework has great potential as rebar, formwork and shoring typically have colours that are quite different from finished concrete (see Figure 3 for example). Furthermore, more detailed object recognition algorithms should be investigated, such as the approach by Ishida et al (2012). Technique A described in this paper, using an expanded 3D model that includes temporary and secondary objects, also needs to be tested and compared.…”