A method of defllectometry based on photogrammetry

A method for deflectometry using photogrammetry has been designed and tested. The system is designed to measure the slopes from a specular or reflec-tive surface, with the design facilitating the optical quality assessment of he-liostats in the department of STERG. The current method operates by attach-ing a planar target to an EOS Mark II camera lens. Using a photogrammetry-based pose estimation, the position and orientation of both the camera and its attached target can then be determined. Using reflective laws allows the cal-culation of normals across the surface. Multiple images with reflections at dif-ferent locations on the reflective surface are stitched together to create a point cloud slope map across the surface. By applying a cubic smoothing spline, we can create an interpolation over the entire surface and construct a full slope map indicating the slope at every location on the reflective surface. Experi-mentation is performed by creating a test surface with a single distorted cor-ner for visual identification and alignment. Experiments are run by performing a system‑slope measurement of the test surface under different system condi-tions, environments, and camera settings. Validation of the results is provided by laser scanning of the surface, which correctly identified the single distorted corner and the slope magnitude and range. Additionally, repeat testing is used to validate the system measurement under the proposition that, because of the system’s measurement method, repeat testing provides validation. Experimen-tation showed a reduction in visual noise with increasing distance from the re-flective surface. System precision is measured using the strongest datasets at two defined distances, <5 m and >10 m. For each distance, three of the best datasets are taken, an average slope map is created, and the error is measured as the distance between each dataset and its average. Accuracy of the system was not determined during the research. Precision is evaluated and presented as the metric of performance. Results indicate higher precision for long‑range data and are used to evaluate the overall precision. The system can produce slope maps with a maximum average precision of ±0.15 mrad RMS error.

Grantor dc:publisher

Stellenbosch : Stellenbosch University

Year dc:date.issued

2026

Author dc:creator

• Winter, Connal Daniel

Advisor dc:contributor.advisor

• Smit, W. J.

Language dc:language.iso

en

Repository record dc:identifier.uri

https://scholar.sun.ac.za/handle/10019.1/135547

OAI identifier oai:identifier

oai:scholar.sun.ac.za:10019.1/135547