QUOTE (wbutler @ Sep 27 2008, 09:54 AM)
And my last post for the moment - The three 2DFFT attributes. I look for the maximal fft value, other than at the origin, and convert to radial coordinates to compute a distance and angle in frequency space. I take the reciprocal of the frequency distance to get a real distance, and I wasn't sure what to do with the angle - I think I can pretty much leave it alone (is there a rotation or something?). I also plot the absolute value of that maximal fft value. This should be related to the strength of the ripples, but will also be related to their regularity. Strong ripples with varying periodicity will spread out the energy over several frequencies, so the maximum will be lower. I think the standard deviation attribute does a better job of this. On the other hand, if both attributes are given to a classifier, maybe it could distinguish strong regular ripples from more chaotic ones.
Great stuff Bill!
My tool works as yet only with the standard deviation, calculating the standard deviation in a 360 degree circle around each pixel, using a range and sample rate which can be configured. I improved a bit on the tool to make it easier to work with and will add some more improvements as I go along, but this standard deviation trick seems to pick out the 'bad' terrain reasonably well. On the other hand, Fourier will get a better indication of the actual 'ripple wavelength', however this only works accurately along the direction of the ripples. I have the feeling that running a FFT along 360 degrees will give you a meaningless answer or at the most something identical to the standard deviation trick.
Click to view attachmentTerrain roughness as indicated by standard deviation goes from green to yellow to orange to red to purple with increasing value (purple is worst). In above example the open bedrock areas are still shown purple, however a lot depends on how you set the resolution, a higher resolution results in a lot more time for calculation, but shows finer details, if I run part of the picture on high resolution the bedrock shows orange or even yellow:
Click to view attachmentWhat I am working on now at the moment is to see if I can make the thing show the general direction of the ripples across the image, and I am trying if I somehow can make a distinction in 'driving direction', in other words big long ripples might be okay to drive as long as you remain in the troughs between them (and as long as these troughs remain open for long distances, no cross-ripples), while they are bad if you need to drive across them. So, as we know the 'required direction' to the destination from every position in the picture, it is possible to see how this translates in terrain classification. Once I get more results of this, I'll show them.
Just to get an impression where we are at the moment, maybe it would be worthwhile if somebody posted one image which we can then all analyze with the various tools, each tool will probably tell a different story but that only makes it more interesting as every trick seems to indicate one specific part and if we put them all together (preferably on a picture showing route already traveled) it is easier to see which tool tells us what. Note it is definitely not a competition to see which trick is 'best' as each has its own specialization, but if we for instance find a route which is voted 'green' in all analyzing methods, that's perfect, the less votes a route gets, the more the 'real' drivers have to take care. Finally, in the end, it all comes down to human eyes, but the tools will help for a first classification of the terrain I guess.
Regards,
Geert.