Date of Award

August 2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical Engineering

First Advisor

Bo Liang

Abstract

This dissertation develops and evaluates a robotic arm-assisted structured-light 3D reconstruction framework for improving reconstruction accuracy and surface coverage under practical experimental conditions. The system integrates camera-projector calibration, Gray-code projection, X/Y decoding, triangulation, MATLAB-based geometric correction, robotic multi-view acquisition, point-cloud alignment, and stitching. First, a tilted-projector structured-light setup was tested with the projector positioned approximately 20 degrees relative to the object plane. This non-ideal configuration introduced global tilt and geometric distortion in the raw point cloud. After correction, the RMSE decreased from 2.8 mm to 0.9 mm, the tilt error decreased from 4.5 degrees to 0.8 degrees, and the dimensional error decreased from 6.2% to 1.7%, confirming improved reconstruction accuracy The correction method was also successfully evaluated across different projector tilt scenarios ranging from 10° to 40°. Second, robotic multi-view reconstruction was applied to a Pharaoh target with an approximate Y-axis length of 200 mm. Since one usable scan covered only about 35 mm, three robot-controlled views were acquired and stitched. The reconstructed Y-span increased from 34.9 mm to 219.7 mm, providing approximately 6.3 times greater coverage than the single-view scan. The valid 3D point count increased from 38,420 to 156,870 points, while the missing-region ratio decreased from 34.6% to 9.8%. Surface-profile analysis showed local overlap mismatches of approximately 0.35 mm and 0.42 mm, with an overall mean stitching mismatch of approximately 1.6 mm. The results demonstrate improved accuracy, point density, surface coverage, and repeatability compared with single-view acquisition. The framework supports future robotic inspection, dimensional measurement, industrial assessment, and hyperspectral 3D imaging.

Download

Share

COinS