Problems encountered

The problems met along the way, from filming underwater to the 3D model, and the tests I ran to solve them. Some pages are still working notes.

While filming

The colour chart is overexposed: at 60 cm the chart fills only a small corner of the frame, and the bright floor around it throws off the exposure. Conclusion: shoot it one stop under.
Colour calibration: the chart filmed on both sides at two depths, to see whether the ambient light changes with depth and how to compensate for it.
Image sharpness: the first images, taken while moving, were blurry, and the point cloud lacked precision. The camera, the water, the lighting and the settings (swimming speed, shutter speed, ISO) all play a part.
The strong current in the canal: how the scooter behaves with the rig (the camera frame) and camera in the current, and how fast the water moves: 35 feet a minute (10.7 m/min), measured by drifting down the canal on June 8, 2024.
Camera rig dimensions: the GoPro’s field of view measured on the floor at 70 cm, and its lens’s angles of view.

Banana effect: a flat 100-foot floor came out curved like a banana after alignment in Metashape. The likely cause: the GoPro’s wide-angle lens distortion. The leads: calibrate the camera, crop the images’ edges, a less wide lens.
Aligning chunks: each zone’s model comes out slightly distorted, which makes them hard to line up with each other. Targets give each one its real size; the distortion isn’t fully solved yet.