C2C x,y,z decomp. output is signed...?

[sunny723d]

Hello, I have Pre ( 26 samples/m2) and Post (56 samples/m2) lidar point clouds of a region that underwent an earthquake. After fine alignment registering (reference dataset is Post), I want to compute a cloud 2 cloud distance. For reference, in C2C differencing, I look at the Post dataset with reference to Pre (so I can see how surface of the earth changed after earthquake).

For C2C differencing:
1. should I subsample the Post dataset so that it better matches the Pre dataset density?

2. When I split into x, y, and z components- there are negative values there. The documentation I have read says that C2C is unsigned... so what are these values representing? As in, if the mean of the z component is negative, does that mean the area subsided?

For M3C2 differencing:
1. should I subsample the Post dataset so that it better matches the Pre dataset density?

2. Is M3C2 better for my purposes?

Thank you.

[daniel]

1. Nope. Always keep as much points as possible, especially for the reference cloud (the more dense the better)

2. It depends. The distance is indeed unsigned (as it's the square root of the sum of squared components: dx, dy and dz). But dx, dy and dz are signed (as they are basically computed as the difference along each dimension of one point to the other). If you consider the Z dimension and you know that the movement is mostly vertical, then indeed you could use this sign information. The best option however is to rely on M3C2 for this kind of analysis.

3. Nope

2. Yes ;)

[sunny723d]

Thank you for the extra info.

1. To be clear:
"C2C absolute distance (z)"
+
"C2C absolute distance (x)"
+
"C2C absolute distance (y)"
=
a vector whose magnitude (length) is represented by "C2C absolute distance"

--> as in, difference vector= x (i hat), y (j hat), z (k hat) with a magnitude of "C2C absolute distance"

2. If that^ is true, then:
a. differencing the point clouds on one side of a known the fault
b. differencing the point clouds on the other side of the fault (treating them as 2 discrete blocks of data),
c. I'll have 2 vectors of movement (ie: difference per side) that I can use to find the vector between them which would be a net slip vector (http://www.geosci.usyd.edu.au/users/pre ... Sld11.html) with calculable magnitude ... does that seem reasonable to you?

[daniel]

1. Well, more simply : dist_3D = sqrt(dx^2 + dy^2 + dz^2) (not sure about the 'hat' thing ;)

2. Well, more than a single vector, you'll have plenty of them. But you may extract the general trend on both sides indeed.