Sensing Underwater #shorts #podcast #science #ai

Hinzugefügt: 2026-05-17

[00:00:00]A lot of the methods we use in robotics for mapping, for example, were developed for imagery we collect in air, and they don't necessarily translate directly to processing these underwater images.

[00:00:13]What are the limitations with mapping or visualizing underwater? On land, we're used to using cameras to get really dense color reconstruction. The process of underwater image formation is actually quite different than on land.

[00:00:24]We have underwater effects like attenuation and backscattering, and these can lead to degradation in images that we see underwater. So, if you've ever taken a picture underwater, maybe diving, you may have seen this that the images that you get back are blue or green in color, and they might have a very strong haze effect across the image. Because of these effects, um, we have a couple of operational challenges.

[00:00:48]One is that we actually can't image very far away from the vehicle that we're working with. So, typically, we do imaging surveys about 2 to 4 m off the seafloor or the target of interest. It also can be, uh, quite challenging to process this data once we get it back on land. A lot of the methods we use in robotics for mapping, for example, were developed for imagery we collect in air, and they don't necessarily translate directly to processing these underwater images. So, this leads to challenges both in uh, data collection and also data processing.