From cameras in the water to a measured result. I build and operate the imaging and data systems that replace manual inspection with consistent, repeatable measurement.
Every ship grows a living layer on its hull. Barnacles, algae, and slime add drag and burn fuel. The coatings that resist them are tested on panels at sea, and for years that was judged by eye. My PhD rebuilt that judgement as a measurement, running as one connected pipeline.
The PhD thesis pipeline
Stage one
The pipeline starts with seeing clearly underwater. I selected and integrated the cameras, optics, and lighting, calibrated them against a submerged target, and built the housings that hold each panel in a fixed position. Three systems have run continuously for more than two years, and have become the standard imaging equipment the test centre relies on.
Stage two
Cameras are only useful when they keep running on their own. Each system is a Raspberry Pi in a sealed enclosure, capturing on a schedule, switching its own lights, and cleaning its own lens. I packaged the software in Docker, deployed it with a GitLab pipeline that ships to each device automatically, and wired everything to PostgreSQL databases and live dashboards.
Stage three
A raw underwater frame is murky, unevenly lit, and full of passing fish and debris. Before anything can be measured, the image has to be cleaned. I built the processing that returns a sharp, evenly lit, geometry corrected picture of each panel, stacking frames to remove drifting particles and fusing exposures for even light.
BeforeAfter
Stage four
A single frame can hold several panels, so before any fouling is measured each panel has to be found and labelled on its own. A detection model locates every panel in the image and gives it a consistent identity, so the right measurement stays attached to the right panel across the whole record.
Stage five
With each panel identified, the models read its surface and measure what has grown on it, separating clean paint from algae and from animal settlement such as barnacles, consistently across every paint colour. The work is openly published with its data and code.
Stage six
The last stage follows panels in motion. On a rotating test rig the same identification and measurement run on panels as they move through the water, which brings the assessment closer to the conditions a real hull experiences and supports dynamic evaluation of how a coating performs in service.
