Endeavour Drive - Drivability analysis Options

[djellison]

http://www.sciencefriday.com/program/archives/200809191

You can listen via NPR, or via one of the web feeds that are listed on the site, but make sure you do listen if you can.

[Juramike]

I think I've got an improvement, but my ignorance of Photoshop image manipulation is holding me back.

The key will be to blur out the dune waves slightly (what blur tool?, pixel size?), then phase shift by 1/2 wavelength (in pixels), then create the differential (differential? multiply? darken? overlay? then blur???)


Anything that doesn't cancel indicates something shifted (I think I need a blur after the differential is created). So signal after differential = bad/caution; signal cancel = good/drivable


Big dunes will have the biggest signal, since the crests will now no longer cancel.
Smaller dunes (with small wavelength) will cancel somewhat have less signal in the differential image.
Flat drivable pavement will totally cancel out, and have no signal and be dark in the differential image.

(And here's the best part)
Flat pavement with big dunes will still have a big signal, since the phase-shifted dunes will not cancel out the pavement response.

I'll try to fiddle with this tonight (since I don't know the right tools or parameters to do this), but does anyone have any suggestions?

-Mike

[jamescanvin]

I think I've got an improvement, but my ignorance of Photoshop image manipulation is holding me back.

If you keep going down this route you might end up with the Fourier technique that I'm using.

Mine effectively boils down to the strength of repeating pattens across parts of the image. Yours is starting to sound like that too.

[RoverDriver]

If you keep going down this route you might end up with the Fourier technique that I'm using.

Mine effectively boils down to the strength of repeating pattens across parts of the image. Yours is starting to sound like that too.

Actually his approach is mor similar to correlation which is not rotationally invariant (as applied in this manner). Since the ripples are elongated features and do not change much direction, it seems to be a good approach and the results are quite good.

An important refinement that FT might be able to generate is the detection of purgatory-like drifts. Purgatory not only had a particular height, but the ridge was somewhat diagonal relative to the prevalent ripple direction. I believe I can see some examples of these in the HiRISE south of VC. Visual detection of these drifts might something that the less image processing savy contributors could do.

This image shows what I'm talking about.