• First structured light 3D scanning for coarse aggregate void content detection. • Establish 3D point cloud system for efficient automated detection. • Design 3D feature extraction, construct a prediction model, and predict void content. • Excellent model performance,meets engineering needs,verifies effectiveness. Void Content of Coarse Aggregate (VCA) is a critical indicator for concrete mix proportion design, directly affecting compactness and mechanical strength. Traditional detection methods suffer from high subjectivity, labor intensity, and long cycles, while existing digital approaches based on 2D images fail to capture 3D spatial morphological features, resulting in limited accuracy. It remains an open scientific question how to to extract 3D morphological features via 3D scanning for in-depth VCA analysis remains an open scientific question. In this study, we propose an automatic VCA detection method using 3D point cloud scanning. First, a binocular structured light device acquires aggregate 3D point cloud data and reconstructs 3D models, from which 3D morphological features are extracted. A decision tree-based adaptive binning method converts particle-level features into fixed-dimensional interval-level statistical features, eliminating dimensional differences among samples. The fused feature matrix of gradation information and processed morphological features is input into a Gradient Boosting Decision Tree (GBDT) to construct a VCA prediction model, optimized by Bayesian hyperparameter tuning. Validated on 80 groups of samples with different gradations, the method achieves a MAE of 0.49%, a RMSE of 0.62%, and a detection accuracy of 98.91%. These data fully verify the feasibility of using 3D structured light scanning for coarse aggregate particles to realize void content detection. As the first proposed fully automated and intelligent detection scheme for VCA, it can significantly shorten the detection time and reduce detection costs, thus providing efficient and accurate technical support for concrete mix proportion design.
Zhang et al. (Wed,) studied this question.