I've whittled down the huge collection of laser-profiled shapes to a small collection that are almost certainly mostly copepods--in particular, finmarks (Calanus finmarchicus). My almost-certainty is based on an anomalously large aggregation at just the depth and size range where we'd expect finmarks to be found. The problem is that they are tiny compared to the resolution of the LOPC, and I would like to convince myself that the blobs I'm seeing really are finmarks.
Here's what the blobs look like in the LOPC data.
There is no way to tell by looking at these (I would argue), that they are copepods, let alone finmarks. (A few of them are probably something else entirely.) In order to help convince myself that these blob shapes are really what I think they are, I've written a script that solves the "forward problem"--i.e. estimating what a copepod would look like if it passed through the laser. I'm sure I'm not the first to have done this, but it only took a few minutes, and it looks pretty neat.
Here's the copepod I start with.
Step one is to outline the shape that passes through the LOPC:
Next I transform the image to binary. This disregards the differing levels of transparency of the animal, but I'm just going for an approximation here, so I'm okay with that.
Finally I pass it through an imaginary LOPC. The amount of dark pixels within the imaginary beam determines the strength of the return. I also use a threshold of about 5% pixel cover. This is what it looks like:
It actually coheres rather nicely with the quasi-silhouettes shown in the first image. Yes, I'm starting to believe, despite the blobular shapes, that these actually are copepods. The next step is to pass the copepod through at different orientations.
Here's what the blobs look like in the LOPC data.
There is no way to tell by looking at these (I would argue), that they are copepods, let alone finmarks. (A few of them are probably something else entirely.) In order to help convince myself that these blob shapes are really what I think they are, I've written a script that solves the "forward problem"--i.e. estimating what a copepod would look like if it passed through the laser. I'm sure I'm not the first to have done this, but it only took a few minutes, and it looks pretty neat.
Here's the copepod I start with.
Step one is to outline the shape that passes through the LOPC:
Next I transform the image to binary. This disregards the differing levels of transparency of the animal, but I'm just going for an approximation here, so I'm okay with that.
Finally I pass it through an imaginary LOPC. The amount of dark pixels within the imaginary beam determines the strength of the return. I also use a threshold of about 5% pixel cover. This is what it looks like:
It actually coheres rather nicely with the quasi-silhouettes shown in the first image. Yes, I'm starting to believe, despite the blobular shapes, that these actually are copepods. The next step is to pass the copepod through at different orientations.
Leave a comment