Optimization of automatic extraction procedure for particles in asphalt mixture towards superior robustness and accuracy

Abstract

Three-dimensional image reconstruction plays a pivotal role in microstructure information extraction and virtual mechanical experiments of asphalt mixture. However, the traditional pixel-based reconstruction algorithm tends to be powerless in processing three-dimensional (3-D) mesostructure. Therefore, this paper aims to develop a robust and accurate algorithm for high-quality aggregate particle extraction. To this end, the global thresholding and three-dimensional mathematical morphological methods were conducted directly based on voxel operation. Then the calculated weight-based gradation was utilized to develop the robustness and accuracy analysis map (RAAM), which assesses and compares the traditional and novel reconstruction algorithms. As a sequence, the voxel-based method had higher accuracy and lower perturbation sensitivity than the existing procedures. Additionally, the computed efficiency was improved dramatically by breaking tight bounds and conservatism. Finally, the inevitable algorithm defects caused by multiple morphological treatments in the traditional method have disappeared in the novel one.