市政技术2025，Vol.43Issue(5)：211-221,11.DOI:10.19922/j.1009-7767.2025.05.211
路用碎砾石颗粒三维形态特征分析与预测
Analyses and Prediction of 3D Morphological Characteristics of Road Gravel Particles
摘要
Abstract
To quickly obtain the 3D shape parameters of road aggregate particles,a 3D shape parameter prediction model was constructed based on the analysis target of 3D shape characteristics of crushed gravel particles.A struc-tured light scanner was used to obtain 3D point cloud models of gravel particles with different sizes,with three-di-mensional shape parameters of elongation,flatness,and concavity;The two-dimensional shape of particles was ob-tained through uniform projection method to analyze the fine length and concavity of particles;The Pearson correla-tion and grey relational analysis methods were used to analyze the correlation between 2D and 3D morphological pa-rameters.Based on the two-dimensional morphological parameters of crushed gravel particles,a 3D morphological parameter prediction model was constructed by random forest model,gradient boosting tree model and stacking model.The effectiveness and influencing factors of the prediction model were analyzed.The research results indicate that the elongation,flatness and concavity of particles with different particle sizes are similar.The shape of particles with a diameter of 13.2～16.0 mm is more closer to spherical,while particles with a diameter of 4.75～9.5 mm tend to be flatter;The overall morphology of crushed stone and gravel particles is similar,but the surface of crushed stone parti-cles is more uneven.All the Pearson correlation between the long axis,short axis,fine length,and elongation of gravel particles is relatively higher(0.45～0.96).The Pearson correlation of gravel particles is higher than that of gravel par-ticles,and the gray correlation of gravel particles is higher than that of gravel particles.The stacking model is supe-rior to the single model in predicting the 3D shape parameters of crushed gravel particles.There is little difference in the prediction accuracy by the three models.But the number of projections has a significant impact on the prediction accuracy.The prediction accuracy is the highest during the 12th projections;Linear and exponential models can't be directly used for predicting the 3D morphological features of single particles,while stacked models have high prediction accuracy for the 3D morphological parameters of single particles.The MSE values for predicting elonga-tion and concavity are about 0.013 and 0.002,significantly lower than linear and exponential models.This research result can provide reference for improving the efficiency and accuracy of particle morphology feature prediction.关键词
颗粒三维形态/碎砾石颗粒/相关性分析/颗粒点云模型/颗粒投影/预测模型Key words
3D morphology of particle/crushed gravel particle/correlation analysis/particle point cloud models/particle projection/prediction model分类
交通工程引用本文复制引用
李宁,沙波,冯宁阳,吴思玥,司伟..路用碎砾石颗粒三维形态特征分析与预测[J].市政技术,2025,43(5):211-221,11.基金项目
国家自然科学基金项目(52308465,52278430) （52308465,52278430）
新疆西域公路建设集团有限责任公司科研项目(220221230614) （220221230614）
