Dear Nicolas,

Have you tried this?

Chang, Y.-C. & Sinha, G. (2007): A visual basic program for ridge axis picking on DEM data using the profile-recognition and polygon-breaking algorithm. Computers and Geosciences, 33(2):229-237. 

It is not very young, but worked on a smaller dataset, I don't know how it would perform on a high resolution laser-scanned DTM. There are some valuable references cited (including our paper, in which we followed a similar approach that Richard has mentioned above :-))... Normally the key point is the skeletonization of your thresholded raster and then the raster-vector conversion.

Kind regards,

Balázs

http://dx.doi.org/ 10.1016/j.cageo.2006.06.007

Hi Nicolas,

You can use spatial tools that use DEMs to create Topographic Position Index (TPI) grids, which can then be classified into landforms, including crests. Here's such a tool that is used in ArcMap. 

It looks like QGIS has similar terrain analysis tools available.

Lara

http://www.arcgis.com/home/item.html?id=b13b3b40fa3c43d4a23a1a09c5fe96b9

Bonjour,

ma réponse est en français.

Vous pouvez extraire automatiquement les crêtes de vos dunes en utilisant bien entendu vos données (DTM) en utilisant le logiciel GlobalMapper (Analysis; find ridges lins)

Bon courage.

Thank you for all your seggestions guys!

Balázs Székely:

Thx for the paper and software you pointed me to. I tried it, as it is opensource, but unfortunatly it doesn't work with LiDAR resolutions. I tried to resample it to coarser datasets but nothing,

Lara M Juliusson:

Thx for the tool! I tried it and it works fine, but it gives me only approximation of the dune crests.

Moussa Ougrin:

Merci bcp pour ton aide Moussa, mais malheureusement j'ai pas ce logiciel. Je vais demander un amis de l'essayer pour moi, mais je pense que je peux trouver le même algorithm qui marche en GlobalMapper pour Qgis ou ArcMap.

Richard Gloaguen:

Thx for the hint, I treated crests as local elevation maxima and I extracted them by running a weighted kernel across the raster and getting the sum of the kernel. I did it with Focal Statistics tool in ArcMap 10.2.

The tricky thing is the weighted kernel: size and power (how much weight  the further pixels in the kernel are considered compared to the inner ones) determine how well the crests are extracted.

I tried with a 5x5 kernel with 1/6 scale factor (power) for the 1m raster and it works fine for my coastal dune setting.

Now, for the 10cm UAV DEM, which has some highly interpolated areas (due to Structure from Motion process), is another story. The kernel must be way bigger (I tried a 101x101) and its power too. I also have to smooth the DEM as much as I can to eliminate all the peaks and bumps that we are not interested to.

Bottomline:

a semi-automated way to calculate sand drift rates from UAV datasets is really challenging, but I won't give up.

Thx again to everyone

Nic

Thank you Antoine for your suggestion!

At the end I elaborated a totally new geospatial technique to extract all the dune crests in a dune field and calculate migration statistics (diplacement and direction) between two LiDAR datasets (2009-2013).

If I find the time, I write it as a script tool for ArcMap 10.3, as it incorporates R scripts, and release it.

Thx for your interest,

best regards

Nic

Hi Nicolas,

You can use of GlobalMapper 17. In Analysis menu while you opened the DEM select generate watershed.

Hi Nicolas,

Did you have the chance to write that script to release? I'm having similar issues trying to identify dune crest from a 0.1 m UAS-SfM derived DEM as well. I have tried a couple of the recommendations without luck and thought I might see. Thanks.