Keywords: lidar

Summary

The purpose of this demo is to show the correct way to configure a Lidar receiver to resolve sub-pixel structure. In this case, the scene is a flat plate with a 10 meter tall box placed on it. The receiver is a single pixel array with a GSD of 1 meter (100 micron pixel through a 1 meter focal length optical path from 10,000 meters). The box is 0.5 x 0.5 x 5 meters and, hence, sub-pixel. Because the box is 10 meters taller than the background, we expect two distinct returns from within this one pixel. In order for DIRSIG to correctly sample the structure within the pixel, the sub-pixel sampling must be greater than 1x1. In this demonstration, a variety of sub-pixel sampling is grid sizes are run to show the user the impact on the returned pulse waveform.

Note
 The sub-pixel sampling demonstrated in this single-pixel example can be applied to multiple-pixel arrays.

Details

To observe the desired effect of multiple returns within a pixel, we will look at the temporal profile contained in the .BIN file produced by DIRSIG. This file contains a spatial by spatial by time cube of the returns for each pulse. The command-line LIDAR .bin analysis tool can be used to easily extract the average temporal profile for this one pixel array:

$ bin_analyze --output_profiles=true demo.bin

The --output_profiles=true option triggers the creation of an ASCII\Text file for each pulse that contains the sum of the photons across the focal plane for each time bin. Since this simulation is only a single pixel and only a single pulse is fired, only one file is created named profile_0.txt.

Note
 If the receiver array contains multiple pixels, the --output_profiles=true option outputs the average waveform for all the pixels in the array.

Important Files

Geometry and Materials

The geometry for this scene is completely defined in the geometry/demo.glist file. The background is defined using the GROUND_PLANE built-in geometry primitive (assigned material ID #1) and the vertically displaced surface is defined using the BOX built-in geometry primitive (assigned material ID #2):

DIRSIG_ODB = 1.0

GROUND_PLANE {
    MATERIAL_ID = 1
}

BOX {
    LOWER_EXTENT = -0.25, -0.25, 0.0
    UPPER_EXTENT = +0.25, +0.25, 5.0
    MATERIAL_IDS = 2
}

The two materials are (ID #1 and #2) are defined in the materials/demo.mat file. Both materials are assigned a diffuse reflectance of 10% using the Ward BRDF model. Separate materials are used so that truth data can reveal the relative proportions of the two objects within the pixel.

Atmosphere

Since this scenario is focused on sub-pixel returns, the Simple radiometry model is used. This means there is no atmospheric absorption.

Instrument

The LIDAR instrument description isn’t notable, except for the sub-pixel sampling. The default (as provided) sub-pixel sampling grid is 11 x 11, but the results of a variety of smaller grid sizes will be shown.

Setup

The following is a step-by-step tutorial that will allow the user to reproduce the results provided below.

  • Run the demo.sim simulation

  • Extract the temporal profile using the bin_analyze tool to extract the average waveform.

  • Plot the resulting profile_0.txt file to observe the two peaks in the pixel returns.

Results

With 11 x 11 Sampling

The instrument setup provided with this demo is A copy of the extracted pixel profile is included in the file returns.txt. The six output columns of the average waveform (profile) file are as follows:

  1. Bin index (starts at 0)

  2. Bin time (seconds after transmit)

  3. Total (polarized and unpolarized) received photons (S0)

  4. S1 received photons (0 for unpolarized return)

  5. S2 received photons (0 for unpolarized return)

  6. S4 received photons (0 for unpolarized return)

A plot of the extracted pulse return waveform is included in the file returns.png. The total (S0) photons are plotted vs. the bin index in the figure below. The waveform shows two peaks: one from the top of the box (near bin 1400) and one from the ground (near bin 1460) Because the GSD of the pixel was 1 x 1 meter and the box is 0.5 x 0.5 meters, we expect to see 75% ground and 25% box in the return. Since both objects have the same reflectance and there is no atmospheric absorption differences for the two ranges, the magnitudes of the two returns are driven by the sub-pixel area of the two objects. In this case the magnitude of the vertical object (at the closer range and smaller time bin) is expected to be 3x smaller than the background it sits on.

Waveform return for the 11 x 11 sub-pixel sampling case.

waveform 11x11

Other Sampling Rates

In order to show the impact of the sub-pixel sampling rate, the same simulation was run with different sub-pixel sampling grid sizes. At the extreme, the 1x1 sampling never samples the ground and only produces the return from the box.

waveform 01x01

waveform 03x03

waveform 05x05

waveform 07x07

waveform 09x09

waveform 11x11