Nice work Bill.
But I doubt if those Anatolia features would give Opportunity much trouble (that is as long as it isn't 'hollow' underneath ).

Yes, I remember a description of a conversation between the scientists and the drivers when they discussed the original Anatolia:

Drivers - "Wait a minute, explain again what you said about 'falling in'".

In any case I think they are difficult to cross and so the rover will have to weave around them.

The color map I chose when doing Victoria does a poor job of showing the Anatolia features - they all ended up green. Maybe I was too optimistic. But they are there in the data, and can be clearly seen in the grayscale image, for example image3a.png.

Does anyone know where/how to get the MOLA pings for Victoria to Endeavor?

My analysis has lingered for technical and time reasons (sigh!) but my DEM method would really spark up if I could see that data (registered to the Hires images?)

any help would be appreciated!

cheers

Merry Christmas to all UMSFers and lurkers!

Go with the HRSC data instead, higher res than the MOLA data.

Doug

Doug,

I have looked at the HRSC data and all I can say is ....Brilliant!

thanks!

Nice job, Bill!

Here is a crop at the edge of Endeavour Crater from your combined analysis image that illustrates some interesting features:
Click to view attachment

On the left there are some filled in craters that have extra large ripples in them. The N-S dune aligment indicates that these were filled in (and bigger dunes heaped up) back when the rest of the N-S aligned Meridiani ripples were set. So it probably happened long ago.

(Why bigger dunes in the filled craters - maybe the sands filling the craters were loose-filled sands and didn't have as many armoring blueberries?)

In the middle are regular N-S trending dunes like we see all over this region.

At the far left is the interior of Endeavour crater. The dunes here are aligned differently, indicating a different wind regime. These are parallel to the recent dune alignment indicated in Sullivan et al., so one interpretation is that these are recently moved dunes (i.e. potential sand traps), [Another interpretation is that the crater interior funneled the local wind direction, but other filled craters with similar alignments in the "floor dunes" just to the N argue agains that possibility.]

Indicated with a red arrow is a small patch of dunes outside the crater with a NE-SW alignment. This is parallel to the recently emplaced dunes. My interpretation is that these are recently emplaced dunes made of fine loose material that escaped the crater floor, only to fetch up against the low ridge (indicated in the Butler model in red). This patch could be a potential deathtrap for Opportunity.

-Mike

I just returned home last week, and have just browsed through the HRSC data also, and indeed it looks very good. There must be a way to combine this with the HiRISE images and find a method to identify for instance the various types of sand in sanddunes. I have some ideas, if I find the time I'll see if I can work them out.

Regards,

Geert.

THEMIS'll give ideas as to the size of sand grains, but unfortunately it's a rather low res... Maybe CRISM data might help?

CRISM data is indeed easy accessible, I have to search to find a way to best match the data we are looking for. Low res is not too much of a problem, as in this case it is not about individual sand dunes but about surface material in general.

Just to give an impression, I originally started with calculating 'roughness' of the surface based on variance in brightness, on the assumption that a lot of variance in brightness will indicate a rough surface. If I make this calculation for the present area Oppy is travelling in, you get the following map:

Click to view attachment

(lightgreen is best, dark red is worst).
As has been mentioned before, this leaves the problem that bedrock area's are shown as 'rough' area's, as the software does not make a distinction between rock and sand.

I can get this distinction if I follow the earlier suggestion of Mike and more or less make an inverted false color image of surface brightness, this SEEMS to make a distinction between rock and sand:

Click to view attachment

Now the 'yellow brick road' nicely stands out as lightgreen and sandy area's and anatolia features turn out red. However, this leaves the problem that nice flat sand area's are not distinguished from rough sanddunes.

So, finally, once again conform Mike's suggestion, I can combine both analyzes, giving the following chart:

Click to view attachment

Now flat sandy area's turn out green, the yellow brick road turns out lightgreen, and sand dunes turn red. The actual mix of how you define the colors is all a bit subjective, just a bit of playing with some parameters, but it is getting somewhere.

What leaves me uncomfortable is that this is all a lot about surface brightness (as defined by the HiRISE filters used), I much rather use a multi spectral scan and see if there is somewhere somehow a correlation between surface composition and 'drivability'. Whenever I find the time I'll just see where this leads me.


Regards,

Geert.


Software tool (requires Windows with .NET 3.5): http://www.navtools.nl/deployment/roverrouter/installnow.htm

Interesting. I did use a variance filter as well and it gave the similar results. Your observation about the intensity playing a role is true. If you think about it, the DC component of the fourier transform is in fact the average brightness of the patch. Since these algorithms are applied to the map projected HiRISE images, which are radiometrically corrected, it is quite possible we are all impressed by classifying mostly image brightness!

I think I need to refresh my memory on co-occurrence matrix texture classification.

Paolo

If I compare the 'sandtrap-analysis' between our 'normal' HiRISE image and the CRISM IR2 data there do seem to be some interesting differences, mainly in the south-eastern portion of the image:

Click to view attachment

Click to view attachment

Topmost is HiRISE, and below CRISM of same area and same scale. Difference in resolution is offcourse clearly visible, but surface brightness as seen by CRISM seems to differ quite a lot mainly in the south and south-east area. Don't know what to make of it, might be just chasing ghosts...

Regarding the technique: I stopped using 'absolute' values some time ago in my tool, what it does is it first calculates average pixel brightness across the complete image, and then creates a dataset of pixelbrightness as a percentage of the average value. Colors are then related to this percentages, so 'red' does not relate to a specific pixel brightness, but instead to a brightness relative to the average brightness of the whole picture. In fact I do the same with the variance, which is also defined as a relative value (so even in an image of a very rough terrain the color green might be seen as 'less rough then average'). This makes it easier to compare images.

Calculating variance-values or a FFT analysis of the CRISM data is useless, as its resolution is too low to catch any sanddunes, I'm just looking if there is any way to catch differences in surface material in relation to "drivability".

Regards,

Geert

Speaking of texture classification:
I've been playing with various texture classification methods (mostly self-written, experimental C-implementations of Co-occurence matrices, Gabor filters, Texture energy based filter banks and others).

The problem I see with most of those methods is that the spatial resolution of the additional information that is gained with the filters (e.g. directional trend of dune crests) will be by far less than the spatial resolution of the original intensity information.

To obtain some significant texture information requires local windows at least 10 pixels wide (or even much more, depending on the method).
Another problem is that of reducing the very high dimensionality of the output of texture filter banks (For example with co-occurence-matrices, we will have one individual band (image) for each direction and each pixel-spacing). Usually there will be dozens (or hundreds) of dimensions that must be mapped to a lower dimensionl space in order to become practical for further processing/visualization.
I have tried various PCA-based and SOM neural networks (Self Organized feature Maps) methods, but many of those require a lot of tuning parameters themselves and therefore further increasing the dimensionality (dgrees-of-freedom) of the original problem.

(So from the point of view of spatial resolution, I think the approach to use additional (low resolution ) spectral bands ( CRISM/THEMIS ) could be more promising than the even lower resolution texture classification approach)

Also, I think as far as the incorporation of textural information in the surface analysis is concerned, we already have a very good coverage by James Canvin's Fourier-based mapping approach. I doubt that any other texture operator could add much additional classifying strength to those maps

Of course! But put this in perspective, if I had to schedule a drive where I need to guarantee the rover position within small margins, I would probably increse the risk factor way too much. 10x10/pix on a HiRISE is only 2.5x2.5m, just above one rover size.



That is correct and that is why I did not try this method yet. Maybe applying feedback from past drives to the classifier would help? I dont know, just thinking out loud.



Besides the much lower resolution of this data, the problem I see is registration of these maps with HiRISE. But this would probably be the method I prefer if I had the means to pursue it.


It might be so, but as I was saying, I would not be surprised if even the FT was somehow mostly influenced by intensity rather than texture.
Ihave to think about it.

Paolo

I've been wondering about this too...

I wonder what the James Canvin FT map would look like using Geert's dataset of %pixel brightness to a moving localized (regional) average?

Would it balance out the natural brightness variation and give a better result?

I've been wondering how to create an artificial "background" gradient, then superimpose (or subtract away) a search for brightness variations with the regular big-dune wavelength.


For the last several weeks Oppy has been driving in an area where the brightness and roughness seems to correlate pretty well.
Assuming Oppy turns to the E, in the next few km, Oppy will go into an area where some of our terrain models diverge big-time.
(This is the region smack dab in the middle of the HiRise images we've been using)

This new region, and points south should make a really good test of the different models.

-Mike

I've tried to do just that, but as of yet such attempts simply break my poor little pc ;-). Variance works okay but my FFT code needs some improvements to make it work with such a large dataset.

I believe the trick of first calculating average brightness and then relating each pixel to this average is certainly worthy of following a bit further to see where it leads us to, but take in mind that it has a disadvantage too: by working with 'relative' data you can only compare data (for drive-analysis) within the same picture, you can no longer state that 'green is maximum drive distances', you can only say that it is 'best possible drive distances in this picture'. But maybe that's what it has been saying all along...



Note that it looks like this region is also where there seems to be a significant difference in HiRISE and CRISM data (see my previous message). Maybe it points to something, maybe it doesn't...

What I'm trying to do at the moment is see if I can create a large, grid-bound, dataset (multi dimensional array) of the victoria-endeavour area incorporating not only HiRISE but also data in other spectral bands. Basically any data should fit as long as its chart-aligned (fits to a lat/lon grid). In the most ideal situation it should even be possible to incorporate navcam/pancam data (convert to polar view, then given a known rover-position add a lat/lon grid). Lots of work but basically it can be done. Once I have this dataset (and probably it will be several datasets for the whole area is much too large) then it should be possible to combine, compare, and analyse data from several different sensors/spectral bands. See where this takes us. Don't expect this to be finished tomorow however ;-).

Regards,

Geert.

ok, I think the "texture classifier" approach should be worth a try
Here is an attempt to create a map (of the area south of Victoria) with a multi-scale bank of texture filters (my own variant of "mini-Gabor" filters, basically designed to capture granularity, directionality, edgeness and other micro-texture features).

Click to view attachment

There are 13 filters applied to each 9x9 pixel moving window ( window-spacing = 2 pixels, with the rest interpolated by bilateral upsampling).
To capture features at multiple scales I decompose the original image in a gaussian pyramid and applay each filter to each pixel at each level of the pyramid,
So altogether we have 13 filters at 7 spatial resolution scales.
I used a PCA-based mapper to reduce this 91-dimensional original feature space into a 2D color space (LAB) that can be used for visually analyzing the resulting texture-map overlayed with the brightness information of the original bw image.

As this is only intended as a first test, I did not try to assign the color-mapping to any meaningful scale in the sense of "dangerous/easy".
The different color just represent the wo most significant dimensions in over-all textural variation as it results from the PCA 91-to-2 dimension reduction.
(coincidentally, a quick Eigenvalue-Analysis shows that two dimensions are already sufficient to capture about 80% of all the variation...)

RESULT:
At a first glance, the usual "terrain classes" seem to be distinguished: "red color tones" = "soft sand", "blue/cyan=nort-south-trending ripples", "orange=east-west oriented ripples", "greenish = bedrock ?", and so on ...
Now, of course, it would be the task of the geologists and rover driver specialists to assign the colors to "meaningful" terrain classes.

For this first test I applied the algorithm only to a reduced 2000x1000 pixel crop of PSP_009141_1780 which took about 200 Seconds running time.
So in principle the approach should be feasible to be applied to full or almost-full resolution imagery as well (thanks to the implementation in good old plain C


If time permits I'm going to download the whole HiRISE-JP2 tonight and try another run at a finer resolution level ....

P.S.: I have no Idea if this is useful at all, just an attempt to see what the incorporation of multiscale texture algorithms could add to the existing brightness based
analysis ...

Wow, this is a pretty good result! I will have to grok this.

Paolo

Here is an new version of the map where I omitted the four highest levels of the gaussian pyramid thus putting a little bit more emphasis at the small scale features:

Click to view attachment


( Unfortunately I don't have nearly as much time for all this fascinating stuff as I wish I had

And yet another version of texture map. This time I used the full three dimensional color space to visualize not only two but three dimensions of the original PCA-projected multi-dimensional feature space (which, however leaves no dimension for the brightness channel anymore

Also, I used a 12x12 pixel moving window and restricted the analysis to the finest resolution pyramid level.

Click to view attachment

Absolutely fantastic!

From a geology point of view, I think it is really neat (and significant) how you've been able to light up the EW (WSW to ENE) trending ripples.
Those are more likely to be the most recently emplaced stuff, and could be a potential problem if it is indicating looser dust.

(And I don't think Oppy has traversed any of these type of areas yet, the closest she got was down near the floor of Endurance)

Is there any way to classify and convert this to a grayscale? It'd be neat to try to see how it correlates with past and future observables.

-Mike

There seems to be a similarity between this analysis and the CRISM IR image.

Both show a much more distinguishable area in the southwest (dark red in CRISM and light purple in yours) and a less accented area in the south and southeast (the green 'passage' in CRISM corresponds with the black area in your analysis). This seems less obvious in other analyze types, hard to say what causes this.

Regards,

Geert

Slowly getting somewhere in trying to combine HiRISE terrain calculations with data from other instruments.

Click to view attachment

Above an overlay of the area immediately south of Victoria with relative infra red brightness as measured by CRISM.

Click to view attachment

Same area, however now with a color overlay of Mafic mineralogy (ir_maf) once again as measured by CRISM.

All as yet very preliminary so tread with caution, due to the various scales and map-projections of the data it is a constant wrestle to convert all data to one and the same chart grid and data format. Positive thing is that once I've got it into the dataset and the correct grid, all further image calculations are very fast and easy.

Regards,

Geert.

Excellent work, Geert ! (I know how time consuming this image processing can be ....)
A step forward in combining as much individual spectral/textural "bands" into one analysis as possible ..

In the same spirit I am now planning to incorprate the Fourier-based Analysis into my texture-analyzer.

Question @ James Canvin:

As I have not followed the details of the former discussion: could you provide the details/parameters of your latest Fourier mapping algorithm ?
(local window size, which components of the FFT used (amplitude, phase) ? pre-processing used, other parameters ?)

Thanks
Bernhard

I'll have to go through my code later to remind myself.

But it is basically the maximum of the Fourier Power Spectrum on ripple scale lengths (~8 - 28 pixels I think) for 1D East-West slices 32 pixels long, each value being an average of ~4 pixels North-South. iirc.

In the search whether chemical surface composition has influence on 'drivability' of terrain, below are compositions from CRISM data and the wellknown HiRISE images of the area around Victoria crater.

Click to view attachment

ferric minerals

Click to view attachment

low-Ca pyroxene

Click to view attachment

high-Ca pyroxene

Click to view attachment

variety of iron minerals

Click to view attachment

olivine or iron phyllosilicates

All of above is in 10 mtr resolution (and that's already a extrapolation from CRISM), no use to try to narrow down the area any further.

If we compare this to the earlier study's of terrain it looks to me like there is clearly some correlation between chemical surface composition and general terrain as measured earlier. When I find the time I see if I can create a similar dataset for the remaining areas covered by MER and let the computer crunch a bit more on correlation factors and such. It's a lot of work but results are interesting.

Regards,

Geert.

This is a fascinating line of inquiry. If you're right then route choices made purely on drivability criteria would result in a degree of observational selection in the suite of minerals encountered by the rover. Of course it could also work the other way, with the routemasters deliberately seeking out chemically distinct areas from the CRISM maps to get the most representative picture. I wonder which it is?

Maybe, but what I'm trying to do goes back to a bit earlier in this thread, when we had a discussion whether sanddunes were associated with certain chemical compositions. If you look at the terrain, there are parts where there are (relatively) big sanddunes, and other area's where there aren't. I can imagine that one of the factors might be the composition of the sand itself, whether or not it easily 'moves'. This is a hypothesis which can be checked by comparing maps of chemical composition with maps of dune-locations, as I have just done.

An other item might be the driving itself, how far do the wheels 'sink' in, how much grip do they have, etc, etc, this might also be related to the chemical composition, with Mike's dataset of driving conditions it might be possible to see if there is a relation there also.

And finally, it has been worrying me a bit that during this whole thread on drivability analysis we have so far only been looking at the HiRISE images, and then mainly at the brightness (no matter if you analyze on variance or FFT or any other combination, the original input is always pixel brightness), so I like to check whether there are other instruments/datasets which can be used to check the results of all the analyse-techniques (Mike has already done a wonderful job in relating the outcome to the actual driving conditions, which allows to check the validity of any hypothesis).

Regards,

Geert.

THEMIS IR-night would be a good start - Sure, it's only 100m/pixel, but we're talking about 15km here.

Correct, THEMIS IR I have also already in the correct format in the dataset, however the "low" res makes it harder to check the correlation with dune size.

Click to view attachment

Above is the Victoria area at 10 mtr res with an overlay of THEMIS Night IR combined with CRISM IR, however this is a preliminary version, I can probably improve a bit further on it by playing a bit more with the settings.

I have also TES data now in the dataset, and even some Viking data although that's even less resolution. I'm still working on the ESA data, there is data but the dataformat is a bit harder to fathom...

Anyway, I think there is quite a lot of material to see whether we can find a correlation with our own previous terrain-models, lots of data to crunch ;-).

Regards,

Geert.

I played a bit with 2D-Fourier-Transforms ...

For the following map I used as texture feature vector the 2D-Fourier-Transform of a 16x16-pixel local window (subdivided in 4x4 bins of the normalized power spectrum )
In order to capture transitions, the whole 16x16 pixel window calculation was done centered at each individual pixel in the original image.
The resulting 16 features are projected via PCA to an (arbitrary) 2D-color space (*)
Interestingly, the second largest Eigenvalue is already ten times smaller than the largest indicating that the intrinsic dimension of the whole feature space is essentially only one.

Processing time was 80 Seconds for 2 Million Pixels.

Click to view attachment

The result (so far) looks quite similar to that obtained by the other texture filters ....
Next I'm going to add other features derived from the local power spectrum and combine them with the other texture filters in my toolbox
(Mini-Gabor, Gaussian Markov, Co-occurence ....)
Furthermore, I should apply the calculations for higher resolution HiRISE maps as well

(*) As usually I did not (yet) attempt to assign any "meaning" to the colors such as "greeen more easy than red". Instead the colors are just arbitrary as output by the mapping algorithm.

And here is the close relative to the Fourier Transform: the Autocorrelation function (ACF) wich shows a similar structure:

Click to view attachment

Click to view attachment

THEMIS Night IR data projected on map of victoria - endeavour range.
Colors are bound to relative value's and range from yellow ('darkest') to green to blue to orange to red (lightest).

It looks like the boundary from the blue/green area to the 'red' area enroute to endeavour corresponds quite exactly to the change from 'dune' terrain to mostly flat planes. If this is correct we might indeed have a tool on hand to predict terrain value's independent from HiRISE coverage.

Click to view attachment

Surface dust levels as measured by TES projected on map of the victoria - endeavour range (100 mtr res). Unfortunately resolution of most TES measurements is too low to be of much help in this, although there do seem to be some trends visible. Colors are once again relative to this area, with same range as previous message.

(The lighter colored square in both images represents the HiRISE coverage of Victoria, just as a position marker)

Regards,

Geert.

Scorecard for Sol 1770:
Click to view attachment

For the UMSF Terrain Models, here are 10 m x 10 m normalized grayscale pixel averages for Oppy's 1770 position:
HiRise (raw): 126.28
HiRise (normalized): 180.86
Malaska 20081003 normalized grayscale Terrain Model: 180.2
Canvin 20081001 normalized grayscale Terrain Model: 181.94
Ontanaya 20081001 normalized grayscale Terrain Model: 110.92
Butler 20081006 normalized grayscale Terrain Model: 132.84
Sassen 20081031 normalized grayscale Terrain Model: 164.17

Scorecard for Sol 1774:
Click to view attachment

For the UMSF Terrain Models, here are 10 m x 10 m normalized grayscale pixel averages for Oppy's 1774 position:
HiRise (raw): 130.78
HiRise (normalized): 202.38
Malaska 20081003 normalized grayscale Terrain Model: 195.59
Canvin 20081001 normalized grayscale Terrain Model: 172.45
Ontanaya 20081001 normalized grayscale Terrain Model: 162.66
Butler 20081006 normalized grayscale Terrain Model: 143.89
Sassen 20081031 normalized grayscale Terrain Model: 199.48

Scorecard for Sol 1776:
Click to view attachment

For the UMSF Terrain Models, here are 10 m x 10 m normalized grayscale pixel averages for Oppy's 1776 position:
HiRise (raw): 130.06
HiRise (normalized): 209.19
Malaska 20081003 normalized grayscale Terrain Model: 185.25
Canvin 20081001 normalized grayscale Terrain Model: 126.66
Ontanaya 20081001 normalized grayscale Terrain Model: 197.64
Butler 20081006 normalized grayscale Terrain Model: 140.59
Sassen 20081031 normalized grayscale Terrain Model: 202.84

Scorecard for Sol 1780:
Click to view attachment


For the UMSF Terrain Models, here are 10 m x 10 m normalized grayscale pixel averages for Oppy's 1780 position:
HiRise (raw): 134.27
HiRise (normalized): 220
Malaska 20081003 normalized grayscale Terrain Model: 174.92
Canvin 20081001 normalized grayscale Terrain Model: 133.25
Ontanaya 20081001 normalized grayscale Terrain Model: 131.69
Butler 20081006 normalized grayscale Terrain Model: 132.78
Sassen 20081031 normalized grayscale Terrain Model: 173.59

Scorecard for Sol 1782:
Click to view attachment

For the UMSF Terrain Models, here are 10 m x 10 m normalized grayscale pixel averages for Oppy's 1782 position:
HiRise (raw): 133.41
HiRise (normalized): 214.88
Malaska 20081003 normalized grayscale Terrain Model: 173.7
Canvin 20081001 normalized grayscale Terrain Model: 143.95
Ontanaya 20081001 normalized grayscale Terrain Model: 134.86
Butler 20081006 normalized grayscale Terrain Model: 149.28
Sassen 20081031 normalized grayscale Terrain Model: 142.69

Scorecard for Sol 1784:
Click to view attachment

For the UMSF Terrain Models, here are 10 m x 10 m normalized grayscale pixel averages for Oppy's 1784 position:
HiRise (raw): 128.06
HiRise (normalized): 182.83
Malaska 20081003 normalized grayscale Terrain Model: 155.88
Canvin 20081001 normalized grayscale Terrain Model: 187.91
Ontanaya 20081001 normalized grayscale Terrain Model: 151.20
Butler 20081006 normalized grayscale Terrain Model: 131.29
Sassen 20081031 normalized grayscale Terrain Model: 218.06

Graph of Sol1713 (Santorini)-->Sol1784 observables and 10x10 m avg. values from the various UMSF Terrain models:
Click to view attachment

-Mike

And here is an EXCEL table with this and a bunch o' derived data: Click to view attachment

Click to view attachment

Relative variance calculated for the present area, this seems to match quite good with IR measurements by CRISM

Click to view attachment

Furthermore there is some match with several chemicals:

Click to view attachment

Relative amount of ferric minerals

Click to view attachment

Iron minerals

Click to view attachment

high Ca pyroxine

I guess above similarities are mostly caused by exposed bedrock on yellow brick road?

Regards,

Geert.

I'd like to think so...
The mineral maps are using IR spectral response (brightness at diagnostic wavelengths) to determine relative abundance of chemicals.

But in the the case of a very subtle mineral variation (i.e. close to overinterpreting spectral noise), they could be "fooled" by brightness variations as well.

(Note how high Ca pyroxene and iron minerals are perfectly anti-correlated, and match perfectly/inversely with IR/RED variance. It would be comforting to see an "outcrop of minerals" that doesn't correlate/anticorrelate to RED brightness/darkness.)

-Mike

As was mentioned much, much, earlier in this thread, it is possible that subtle regional brightness artifacts are significantly influencing all the terrain models. It would be neat to remove/subtract the regional brightness variations to leave just the local brightness variations due to the ripple patterns. The remaining patterns could then be used for FT analysis, differential shift analysis, and pattern recognition.

Assuming every pixel in the HiRise image is a result of dune top brightness (A) + background (B), then the dune trough pixel value could be used to create the background (B) layer. (Note that the minimum value would be best to make the background layer, not the average value).

Click to view attachment

The trick will be to find the best way of using Photoshop to create a minimum brightness layer that then gets “blurred” to fill in the gaps for eventual image subtraction. Any advice?


If anybody want to try their hand at this, I’ve placed a normalized HiRise image of the S Victoria region here as a JPEG file (12.5% full HiRise or 2 m/pixel scale): http://www.flickr.com/photos/31678681@N07/3249065487/
A 30 Mb TIFF file of the same image is available here: http://www.speedyshare.com/643201551.html

I think indeed all terrain calculations are more or less influenced by brightness variations, if you just calculate a false color image of the brightness you already get results which closely match terrain calculations which is weird. This however does not say we are only calculating 'noise', as you have proved all to clearly there is a nice match between actual rover driving conditions and predictions from our terrain models. Furthermore there is a good match with THEMIS night IR measurements as I have shown earlier, THEMIS seems to show very nicely were the dune-area's are and this closely matches our predictions.

I think the problem with filtering is that the dune's aren't nice sinusoid and their 'wavelength' as well as their direction variates a lot, so trying to filter on one particular wavelength/direction results in the loss of a lot of information. Furthermore, the bedrock-area's as well as crater-edges and such causes a lot of 'disturbance' as they are also exceeding the average brightness. Dunes on exposed bedrock are different from dunes on sand, or at least they are to our filters.

I am not working with photoshop, as I convert the images to a multi-dimensional matrix and then work with software on the data itself which gives me the most flexibility in what I wish to calculate and show, but what I'm trying to do at present is see if I can let the software recognize certain 'fingerprints' of variations (for instance caused by bedrock, or craters, or dunes, etc), and work from there onwards.

There is a lot more to the terrain calculations then only showing the dune's, by the time we have perfected our 'dune-model' we might be already outside that area!

Regards,

Geert.

A first crude attempt to reduce the effects of bedrock on the terrain analysis

Click to view attachment

Basically a calculation of variance on relative brightness values (differences from average brightness) where after a second filter finds area's which have a considerably higher brightness then there surroundings over a larger area and mark these area's green, leaving only dunes marked blue or red.

It is still not perfect and requires quite a bit of playing with settings before you get a 'reasonable' picture, so it remains work in progress.

I'm working on a second filter to spot craters, which looks hopeful providing the crater is more or less 'round'...

Regards,

Geert.

This all hurts my head - It's so cool and esoteric!
Tell me, Geert, is your 'bedrock-identifying filter' able to recognize that the bedrock is predominantly very bright 'evaporite', crisscrossed by quite narrow (less than 2 or 3 cm wide) sand-dark (or shadow-dark) lines. It seems to me that the close juxtaposition of sand-dark lines with evaporite-bright areas that constitute the bulk of the surface area (I'm guesstimating upward of 70%), will only be found on exposed bedrock, never on totally sandy areas. YMMV.

I can't identify lines of 2-3 cm wide on HiRISE images which have a resolution of 25 cm...

But you're absolutely right, that seems the best description of 'bedrock' and it might work if you use pancam or navcam images (I can convert pancam/navcam to polar view and then project on the HiRISE map, but then you have only a very small area to work in).

Presently I'm just identifying area's which have a very bright signature and extend over several pixels in all directions (small bright lines might be dunes). Hard thing is to distinguish from crater walls (which are also partly exposed bedrock, however often not 'nice' for driving), but I'll work on craters later ;-).

Regards,

Geert.

That's the crazy part of it, Geert: The cracks in the bedrock are too narrow to be resolved by HiRISE! Just like the shadow of the camera mast on Oppy was too narrow to be seen at Victoria.
But I can see them all in the highest resolution HiRise views!! So what gives??

Actually you are not seeing the cracks Shaka. You are seeing subtle discontinuities in the textures on either side of the cracks. The human mind is capable of discerning minor contrasting details in a relative context that it is unable to discern in a homogeneous observation.