Date Approved

8-10-2026

Embargo Period

8-10-2026

Document Type

Thesis

Degree Name

M.S. Civil Engineering

Department

Civil and Environmental Engineering

College

Henry M. Rowan College of Engineering

Advisor

Adriana Trias Blanco, Ph.D.

Committee Member 1

William T. Riddell, Ph.D.

Committee Member 2

Huaxia Wang, Ph.D.

Committee Member 3

Jess Everett, Ph.D.

Committee Member 4

Samantha N. Valentine

Keywords

Building Information Modeling;Computer Vision;LiDAR;Machine Learning;Remote Sensing;Semantic Segmentation

Disciplines

Geography | Remote Sensing | Social and Behavioral Sciences

Abstract

This thesis presents a BIM-oriented methodology for the automated classification, measurement, and segmentation of structural trusses from point cloud data. A real truss point cloud and a complementary synthetic dataset derived from CADBIM Revit families were used to develop and evaluate the workflow. Classification was performed using a rule-based Python algorithm based on principal component analysis, side-view feature extraction, and family-specific geometric decision logic, while a separate family-aware script was used to extract key truss dimensions. Semantic segmentation was evaluated using a PointTransformerV3 model to separate truss and non-truss points in mixed point cloud scenes. The results demonstrated strong performance across all major tasks, indicating that structural trusses can be effectively identified, measured, and segmented from point cloud data. Overall, the study provides a strong foundation for future BIM- oriented automation workflows involving structural point cloud analysis.

Download

Share

COinS