Can't distinguish objects within heteregeneous soil (Peplinski method)

[Jack H]

Hi all and the gprMax team,

I've been using gprMax just for a little while as a data scientist (read: little to no experience with geology/geophysics!) and have been generating randomised Bscans with several buried objects. I've had success locating buried boxes and cylinders in basic simulations of clay/sand/loam, etc (see an example in the objects_in_clay_* files below), with the geometry and objects all falling where I expect them to be and being rather obvious.

However, when I change the basic box to a fractal box with the soil_peplinski characteristics, using:
#soil_peplinski: 0.2 0.8 1.09 2.66 0.001 0.25 my_soil

#fractal_box: 0 0 0 length tx_rx_ypos width 1.5 1 1 1 50 my_soil my_soil_box

There is almost no return, and what is there is very distorted and low-magnitude (see the attached objects_in_soil* files). Am I missing in the application of the above two commands or am I using them incorrectly? Or, are the signals behaving as expected and are just too small to observe in this way (even with the squashed range of the imshow)?

Also, I am very interested in the use of your landmine/IED objects, but I see that they are not yet currently available. Is a release date on the horizon? 

Thanks in advance for any help you can supply,

Jack

[Jack H]

(Sorry, could not post with the '.in' files for some reason, so I've had to attach them separately):
#-----------------------------------------------------------------------------------------------------------------------------------------------------------------------

#title: B-scan from objects buried in clay

#domain: 7.5 1.0 0.002

##c/(max_f*sqrt(rperm)) = 3e8/(3e9*sqrt(6)) = 0.0408, 3e8/(12e8*sqrt(6)) = 0.102

#dx_dy_dz: 0.004 0.004 0.002 

## time window is calculated by: 2*d/(c/sqrt(rperm)), where d=distance to target, c = 3e8, rperm obvious

#time_window: 1.28e-08 

##object materials (minus pec)

#material: 4.0 1e-8 1 0 pvc

#material: 3.0 1e-4 1 0 wood

#material: 1.01 5e-15 1 0 air

#geometry_view: 0.0 0.0 0.0 7.5 1.0 0.002 0.06 0.02 0.002 Bscan_cR38M967J2_geom.vtk n

#waveform: ricker 1 1000000000.0 my_ricker

#hertzian_dipole: z 0.06 0.95 0 my_ricker

#rx: 0.15 0.95 0

#src_steps: 0.02 0 0

#rx_steps: 0.02 0 0

#material: 12.5 0.01 1 0 clay

#box: 0.0 0.0 0.0 7.5 0.95 0.002 clay

#box: 3.710 0.180 0 4.050 0.520 0.002 wood

#box: 6.290 0.430 0 6.490 0.630 0.002 air

#box: 5.490 0.110 0 6.210 0.830 0.002 pec

#box: 0.370 0.050 0 1.150 0.830 0.002 pec

#cylinder: 3.270 0.690 0 3.270 0.690 0.002 0.170 wood

#box: 6.570 0.010 0 7.150 0.590 0.002 air

#cylinder: 1.760 0.490 0 1.760 0.490 0.002 0.350 air

#cylinder: 4.870 0.470 0 4.870 0.470 0.002 0.310 pvc

#cylinder: 2.830 0.370 0 2.830 0.370 0.002 0.300 wood

#cylinder: 7.210 0.510 0 7.210 0.510 0.002 0.110 air

#-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
#-----------------------------------------------------------------------------------------------------------------------------------------------------------------------

#title: B-scan from objects buried in soil

#domain: 2.5 1.2 0.002

##c/(max_f*sqrt(rperm)) = 3e8/(3e9*sqrt(6)) = 0.0408, 3e8/(12e8*sqrt(6)) = 0.102

#dx_dy_dz: 0.002 0.002 0.002 

## time window is calculated by: 2*d/(c/sqrt(rperm)), where d=distance to target, c = 3e8, rperm obvious

#time_window: 2.4e-08 

##object materials (minus pec)

#material: 4.0 1e-8 1 0 pvc

#material: 3.0 1e-4 1 0 wood

#material: 1.01 5e-15 1 0 air

#geometry_view: 0.0 0.0 0.0 2.5 1.2 0.002 0.06 0.02 0.002 Bscan_E6TtTnHEBO_geom.vtk n

#waveform: ricker 1 1000000000.0 my_ricker

#hertzian_dipole: z 0.06 1.0999999999999999 0 my_ricker

#rx: 0.15 1.0999999999999999 0

#src_steps: 0.01 0 0

#rx_steps: 0.01 0 0

#soil_peplinski: 0.2 0.8 1.09 2.66 0.001 0.25 my_soil

#fractal_box: 0 0 0 2.5 1.0999999999999999 0.002 1.5 1 1 1 50 my_soil my_soil_box

#box: 1.080 0.100 0 1.560 0.580 0.002 air

#cylinder: 0.730 0.700 0 0.730 0.700 0.002 0.310 wood

#-----------------------------------------------------------------------------------------------------------------------------------------------------------------------

[Craig Warren]

Dear Jack,

I suspect you have not set the seeding for the fractal box command - see parameter i2 in the docs at http://docs.gprmax.com/en/latest/input.html#fractal-box The #fractal_box command and #add_surface_roughness command both use a random number generator to generate the fractals. If you don't seed the random generator (by specifying i2 with any number) then you will get slightly different fractals ever time you run a model (A-scan) - which is not what you want when you are creating a B-scan, you want the soil distribution to be the same for all the A-scans.

Kind regards,

Craig

[Jack H]

Thanks for that Craig! I seem to be getting a reasonable return now (almost nothing from my "wood"/"pvc" objects but I guess that is correct). I think I understand how seeding affects the Ascans now,  since a lot of my confusion was stemming from the fact that my paraview was looking fine (but was not representative of the soil distribution at each Ascan). 

My current issue that I am facing the issue is that I'm unsure about how to calculate the distance of the new objects since the permittivity is no longer static. At the moment I'm lazily assigning the max permittivity of the soil, thus overestimating the depth of the objects in general. My guess is that depth calculation is only an estimation in real-world scenarios, and is likely mirrored here. Again: not a geophysicist. 

Also, sorry to repeat myself, but is there any news on the buried landmine objects? I understand if you are awaiting a publication on two : ) , but I just haven't seen any recent updates.

Thanks again,

Jack

[Jack H]

Just for anyone's interest, here is my latest Bscan in file and output png (notice the 4 objects are 2 PEC (1 off screen...), 1 pvc, and 1 wood object.)

--------------------------------------------------------------------------------------------------------------------------------------------

#title: B-scan from a metal cylinder buried in a dielectric half-space

#domain: 2.5 1.2 0.002

##c/(max_f*sqrt(rperm)) = 3e8/(3e9*sqrt(6)) = 0.0408, 3e8/(12e8*sqrt(6)) = 0.102

#dx_dy_dz: 0.002 0.002 0.002 

## time window is calculated by: 2*d/(c/sqrt(rperm)), where d=distance to target, c = 3e8, rperm obvious

#time_window: 2.4e-08 

##object materials (minus pec)

#material: 4.0 1e-8 1 0 pvc

#material: 3.0 1e-4 1 0 wood

#material: 1.01 5e-15 1 0 air

#geometry_view: 0.0 0.0 0.0 2.5 1.2 0.002 0.06 0.02 0.002 Bscan_31CRvgI1OG_geom.vtk n

#waveform: ricker 1 1000000000.0 my_ricker

#hertzian_dipole: z 0.06 1.1 0 my_ricker

#rx: 0.15 1.1 0

#src_steps: 0.01 0 0

#rx_steps: 0.01 0 0

#soil_peplinski: 0.2 0.8 1.09 2.66 0.001 0.25 my_soil

#fractal_box: 0 0 0 2.5 1.1 0.002 1.5 1 1 1 50 my_soil my_soil_box 14482

#cylinder: 1.290 0.540 0 1.290 0.540 0.002 0.340 wood

#box: 2.070 0.170 0 2.470 0.570 0.002 pec

#cylinder: 0.460 0.440 0 0.460 0.440 0.002 0.340 pvc

#cylinder: 1.810 0.800 0 1.810 0.800 0.002 0.220 pec

[Craig Warren]

Dear Jack,

Your assumptions are correct - you are very limited in calculating exact target depths when the soil is a heterogeneous distribution of different materials with different dielectric values, i.e. like the Peplinski/Dobson model. You could use an average velocity value based on the range of dielectric values, but doesn't account for what volume of that material is used in the distribution.

The plan is release the landmine models with the next version of gprMax. This has been delayed because I have been too busy with other work commitments to be able to get the version ready for release. At the moment that is looking like late Spring or early summer.

Kind regards,

Craig

[Jack H]

Hi Craig,

Glad to hear about the landmines still having a planned release. For my work, I guess metal cylinders and boxes will do until then! :) 

Also, just a quick thanks to you and your lab for making such a useful tool freely available. It's been invaluable in helping me understand GPR technology and its fundamentals (to my extent, at least).  

Regards,

Jack