Can LIDAR by itself, tell the difference between rock (large and small) and soil? I'm under the assumption that the answer is no. However, I wanted to check here with the experts as I was told by someone else (not an expert) that it is possible.

If it is not possible and you know of what sensor data for me to look into, that would be very helpful.

  • 1
    Did you google "lidar and sediment classification"?
    – risail
    Feb 26, 2016 at 1:42
  • Is there a clear difference between the two? At what point does rock and dirt become soil with rocks in it?
    – tomfumb
    Mar 1, 2016 at 17:27
  • Good point tomfumb. I should of been more clear. I guess initially, I just want to tell the difference between soil (or soil with small rocks in it) and mostly rocky soil. Imagine shooting a bullet into the dirt. Would it absorb into the ground? Or ricochet off? I want to try and find soil that would absorb a bullet.
    – DHammer
    Mar 2, 2016 at 18:32

2 Answers 2


In terms of something akin to a spectral signature, the only way would be through the return intensity values, which are rarely calibrated. Unfortunately, there is really nothing expected in the characteristics of the return intensity that would separate rock and soil, the answer really is that this is not a likely outcome.

Now, if you used surface texture to identify areas of homogeneous surface in contrast to heterogeneous as a indication of large rocks verses bare soil, that may yield some type of usable result.

With a spectral sensor, this would likely be two separable classes.

  • You Articulated the point much better than my original comment. Remote sensing Sediment classification is an ongoing research topic that can't be answered in a q&a forum.
    – risail
    Feb 26, 2016 at 4:39
  • Yeah, this is pretty true. I'd add that there could be some physical differences as well as spectral differences — soil, as a rule of thumb, could be considered rougher than a rock surface. So, you could use some means to classify points by their roughness, either through some sort of neighborhood roughness estimate or, if you have the full waveforms available, by examining the waveforms themselves. However, this seems like a relatively untested approach — you'd be treading some new ground, as far as I know. Feb 26, 2016 at 18:04
  • 1
    @Pete Gadomski in this case we are using texture and roughness interchangeably. I do not believe that you would need the waveforms but could do this on an interpolated all return or last return surface using discrete return data. If you make a stretch, this is not entirely new ground. McKean et al., (2008) used blue-green lidar and wavelet analysis to classify channel substraight types, separating out cobble, to identify salmon red habitats. Several years ago, I developed an algorithm, for the DOD, to identify potential unexploded ordnance using a multi-scale wavelet analysis. Feb 26, 2016 at 18:23
  • This is all very helpful and gives me a starting point. Thank you for your help.
    – DHammer
    Feb 27, 2016 at 16:36
  • @JeffreyEvans thanks for the reference. It appears McKean et al. applied the wavelet decomposition on an elevation profile derived from gridded returns, not on the LiDAR return data themselves — a bit of a different approach than what I was proposing. I was wondering if you could try to make a roughness guess from the "cleanliness" of the full waveform return, under the assumption that a rougher surface would have a wider ("messier") return. My comment was mostly designed to point out that spectral information wasn't the only option, but that spatial information could be used as well. Feb 29, 2016 at 14:23

You also may check out CANUPO, which makes use of multi-scale dimensionality...no intensity values required for this one. Works best when build signatures are highly differentiating. So this might work rather good when you want to differentiate between soil (flat) and rock (rugged). CANUPO is contained in Cloudcompare; but there is also a stand-alone version to automate things. Plus: it's entirely free!

€dit: Oh, you mean the lidar device itself? Then the answer is no. All the Lidar sensor does is recording the distances by measuring the time for a signal to return. Based on the sensor used you may gain some additional information such as intensity or echo/wavelength. But for any classification you need some appropriate software.

  • Interesting stuff
    – huckfinn
    Feb 27, 2016 at 23:39
  • This software, and associated methodology, looks promising. I forwarded it along to a few colleagues that are still involved with lidar research. Feb 29, 2016 at 16:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.