Originally Posted by
mAAine_flyer
Not quite correct. A panchromatic sensor is a single channel detector sensitive to EM radiation (i.e., light) over a broad range of wavelengths (usually corresponding to the visible part of the spectrum). It produces a grayscale image. For an 8-bit data product (typical for satellite imaging systems), each pixel has 256 possible shades of gray (0=black, 255=white).
Unfortunately, that makes it even worse. While the visible spectrum makes it nice for us humans to look at things, trying to detect something in the water will be made more difficult because of the blue and green in the mix.
Originally Posted by
mAAine_flyer
True. But imaging systems with pan and multispectral sensors offer the potential for pan-sharpening, i.e., fusing the two data products and essentially using the panchromatic image to "sharpen" the resolution of the spectral bands.
The way we did things in the project I mentioned in a previous post was to use clustering to bring out the likely reflectivity of various land cover we were trying to identify.
I'd imagine the same sort of process could be used in what you mention. Not so much to 'sharpen' the resolution of the bands, but to assign a specific land cover type to a specific cluster.
Now, with a view over the ocean, there is only two land covers to try to find: Water, and not water. Water will absorb almost all IR light, so almost anything that shows up in the IR with high reflectivity will be a floating object. Then again, a cresting wave may show up with a good deal of reflectivity as well.
In any case, I think it's a difficult task.