Automated rebar diameter classification using point cloud data based machine learning
Introduction
Inspecting rebar diameters and rebar spacing is important for fabricators and site engineers to check dimensional compliance with the design model during the manufacturing and construction stages. This is because the bearing capacity of the reinforced concrete structures is dictated by the size and position of the rebar. Therefore, rebar of the correct size should be installed in the correction position, as determined by the blueprints, to ensure the structural integrity of the reinforced concrete structure. In this regard, dimensional inspection of rebars is conducted primarily by qualified workers to detect any abnormalities related to rebar diameter and rebar spacing using measurement tapes. However, this is a time-consuming and labor-intensive task, so there is an urgent need for automated rebar diameter and rebar spacing measurement that can save time and increase the reliability of the inspection.
Thanks to improvements in 3D sensing technology, several studies have been conducted into dimensional inspection of prefabricated reinforced concrete components such as precast slabs [[25], [26], [27],29,49] and precast girders [51] over the past decade. However, there have been relatively few studies [1,21,34] into rebar inspection. Recently, the authors' group proposed a technique that estimates the dimensions of rebar and formwork using the RANdom SAmple Consensus (RANSAC) [16] based on terrestrial laser scanning (TLS) approach. However, the prior study assumes rebar diameters are known prior to use as the input parameters of the RANSAC algorithm, making the algorithms unreliable due to the manual input. To fully automate the process, rebar diameters will need to be predicted from raw rebar scan data without any manual input. However, determining rebar diameters is a challenging task due to the small size of the rebar and irregular shapes of the rebar surfaces. In addition, diameter prediction in the current circle fitting methods remains a major challenge in the presence of noise, outliers, distortion, and missing boundaries in the unregistered scan data [28]. In order to tackle these technical issues, this paper aims to develop a TLS-based rebar diameter classification technique that automatically classifies rebar diameters using a machine learning approach. In this study, a new concept of density-based machine model is proposed and validation tests are conducted to demonstrate the applicability of the proposed rebar diameter prediction technique. The uniqueness of the study are (1) the development of a rebar diameter classification technique for the first time; and (2) successful applicability validation of the proposed technique through various tests including comparison tests with traditional methods.
This paper is organized as follows. Related background of the study and state-of-the-art studies is detailed in Section 2, followed by explanation of the proposed method and its procedure in Section 3. Next, validation tests and results are presented and interpreted comprehensively in Section 4. The primary factors affecting the result of the proposed method are discussed in Section 5. Finally, this study is concluded with a summary and future work.
Section snippets
Overview of circle fitting methods
There are three main approaches in classical circle fitting, which are 1) geometric, 2) algebraic, and 3) robust fitting. Geometric fitting minimizes the sum of the squared geometric (orthogonal) distance from the estimated circle to the given data points. According to [8], geometric fit is commonly regarded as the most accurate, but it is implemented by iterative schemes that are computationally intensive and subject to occasional divergence. Another limitation of geometric fitting is that its
Methodology
Fig. 1 shows the workflow of the proposed rebar diameter classification, which consists of three stages: 1) training, 2) prediction and 3) rebar spacing estimation. The training stage is composed of five sub-steps, including data collection, data pre-processing, feature extraction, feature selection and machine learning model selection. In the prediction stage, new data sets are prepared and predicted using the learning model chosen in the training stage. Note that there is no step involving
Data collection for training
Fig. 5 shows the two different data collection set-ups. A phase-shift TLS, FARO M70, with a measurement accuracy of ±3 mm in a range of 0.6 m to 70 m and a measurement rate of up to 488,000 points/s, was used for data acquisition. For the first set-up as shown in Fig. 5(a), 14 individual rebars with 7 different diameters (D10-D40), as shown in Fig. 5(c), and a standard length of 3 m, were scanned. Note that the rebar layers were scanned at 14 random scan positions with two different scan
Discussion
To further identify the effectiveness of the proposed method, further investigation into the three aspects, which are 1) accuracy comparison between the traditional machine learning approach and the proposed DBM approach; 2) optimal number of density group; and 3) recommendation of scan density for performing rebar diameter classification, was conducted.
Conclusion and future work
This study presents a new TLS-based approach that automates the classification of rebar diameters using machine learning in order to enable accurate rebar spacing inspection. In this study, a new methodology named Density based Modeling (DBM) is proposed to improve classification accuracy. Experimental tests on laboratory specimens with rebars of seven different diameters (D10-D40) were conducted and the results show that the prediction accuracy for large rebar diameter group (D25-D40) was up
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
The first author would like to acknowledge that this research was supported by two fundings (1) the National Research Foundation of Korea (NRF) grant funded by the Korea Government (MSIP) (No. 2018R1A5A1025137) and (2) the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2016R1A6A3A03010355).
References (52)
- et al.
Real-time drill monitoring and control using building information models augmented with 3D imaging data
Autom. Constr.
(2013) Generalizing the Hough transform to detect arbitrary shapes
Pattern Recogn.
(1981)- et al.
Classification of sensor independent point cloud data of building objects using random forests
J. Build. Eng.
(2019) - et al.
Application of a new in situ 3D laser scanner to study the scale effect on the rock joint surface roughness
Int. J. Rock Mech. Min. Sci.
(2004) - et al.
Discrimination between marls and limestones using intensity data from terrestrial laser scanner
ISPRS J. Photogramm. Remote Sens.
(2009) - et al.
A framework for dimensional and surface quality assessment of precast concrete elements using BIM and 3D laser scanning
Autom. Constr.
(2015) - et al.
Automated dimensional quality assessment of precast concrete panels using terrestrial laser scanning
Autom. Constr.
(2014) - et al.
Automated dimensional quality assessment for formwork and rebar of reinforced concrete components using 3D point cloud data
Autom. Constr.
(2020) - et al.
Automated dimensional quality assurance of full-scale precast concrete elements using laser scanning and BIM
Autom. Constr.
(2016) - et al.
Pipe Radius Estimation Using Kinect Range Cameras
(2019)
Hyperspheres and hyperplanes fitted seamlessly by algebraic constrained total least-squares
Linear Algebra Appl.
Robust statistical approaches for local planar surface fitting in 3D laser scanning data
ISPRS J. Photogramm. Remote Sens.
Robust statistical approaches for circle fitting in laser scanning three-dimensional point cloud data
Pattern Recogn.
Outlier detection and robust normal-curvature estimation in mobile laser scanning 3D point cloud data
Pattern Recogn.
Automated defect detection and classification in ashlar masonry walls using machine learning
Autom. Constr.
Using symmetry in robust model fitting
Pattern Recogn. Lett.
Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning
Autom. Constr.
Principal component analysis
Chemom. Intell. Lab. Syst.
Optimal placement of precast bridge deck slabs with respect to precast girders using 3D laser scanning
Autom. Constr.
Error analysis for circle fitting algorithms
Electron. J. Stat.
Some theory for Fisher's linear discriminant function,naive Bayes', and some alternatives when there are many more variables than observations
Bernoulli
Classification and Regression Trees
Quantitative Evaluation of CHT and GHT for Column Detection Under Different Conditions of Data Quality
Circular and Linear Regression: Fitting Circles and Lines by Least Squares
Least squares fitting of circles
J. Math. Imag. Vision
Support-vector networks
Mach. Learn.
Cited by (33)
Smart rebar progress monitoring using 3D point cloud model
2024, Expert Systems with ApplicationsAutonomous dimensional inspection and issue tracking of rebar using semantically enriched 3D models
2024, Automation in ConstructionAutomatic spacing inspection of rebar spacers on reinforcement skeletons using vision-based deep learning and computational geometry
2023, Journal of Building EngineeringApplicability of smart construction technology: Prioritization and future research directions
2023, Automation in ConstructionPixel-level classification for bridge deck rebar detection and localization using multi-stage deep encoder-decoder network
2023, Developments in the Built Environment