3D Image Analysis Using Deep Learning for Size and Shape Characterization of Stockpile Riprap Aggregates—Phase 2

Haohang Huang, Erol Tutumluer, Jiayi Luo, Kelin Ding, Issam Qamhia, and John M. Hart

Permalink

https://hdl.handle.net/2142/114476 Copy

Description

Title

3D Image Analysis Using Deep Learning for Size and Shape Characterization of Stockpile Riprap Aggregates—Phase 2

Author(s)

Haohang Huang, Erol Tutumluer, Jiayi Luo, Kelin Ding, Issam Qamhia, and John M. Hart

Issue Date

2022-09

Keyword(s)

• Riprap
• Aggregate Stockpile
• Field Imaging
• Computer Vision
• Deep Learning
• 3D Reconstruction
• 3D Segmentation
• Size and Shape Analyses

Abstract

Riprap rock and aggregates are extensively used in structural, transportation, geotechnical, and hydraulic engineering applications. Field determination of morphological properties of aggregates such as size and shape can greatly facilitate the quality assurance/quality control (QA/QC) process for proper aggregate material selection and engineering use. Many aggregate imaging approaches have been developed to characterize the size and morphology of individual aggregates by computer vision. However, 3D field characterization of aggregate particle morphology is challenging both during the quarry production process and at construction sites, particularly for aggregates in stockpile form. This research study presents a 3D reconstruction-segmentation-completion approach based on deep learning techniques by combining three developed research components: field 3D reconstruction procedures, 3D stockpile instance segmentation, and 3D shape completion. The approach was designed to reconstruct aggregate stockpiles from multi-view images, segment the stockpile into individual instances, and predict the unseen side of each instance (particle) based on the partial visible shapes. Based on the dataset constructed from individual aggregate models, a state-of-the-art 3D instance segmentation network and a 3D shape completion network were implemented and trained, respectively. The application of the integrated approach was demonstrated on re-engineered stockpiles and field stockpiles. The validation of results using ground-truth measurements showed satisfactory algorithm performance in capturing and predicting the unseen sides of aggregates. The algorithms are integrated into a software application with a user-friendly graphical user interface. Based on the findings of this study, this stockpile aggregate analysis approach is envisioned to provide efficient field evaluation of aggregate stockpiles by offering convenient and reliable solutions for on-site QA/QC tasks of riprap rock and aggregate stockpiles.

Publisher

Illinois Center for Transportation/Illinois Department of Transportation

Series/Report Name or Number

FHWA-ICT-22-013

Type of Resource

text

Language

en

Handle URL

https://hdl.handle.net/2142/114476

Sponsor(s)/Grant Number(s)

IDOT-R27-214

Copyright and License Information

No restrictions. This document is available through the National Technical Information Service, Springfield, VA 22161.

Owning Collections