LiDAR Series: Everything you need to know before starting a project with LiDAR data.

If you want to start your project in autonomous vehicles, tsunami prediction, astronomy, or 97 more areas, you will have to use something…

Hold tight. We are meeting LiDAR.

Just like RaDAR (Radio Detection And Ranging) uses radio waves, SoNAR (Sound Navigation And Ranging) uses sound propagation, LiDAR (Light Detection And Ranging) uses light to determine ranges (variable distance) by targeting an object with the laser pulses and measuring the time for the reflected light to return to the receiver.

Depending on the purpose, you can use either local ground, aircraft, or space collected data. Airborne LiDAR (which is the one that is collected using airplanes or drones) is nowadays the most precise and accurate method of creating high-resolution digital (3-D point cloud) elevation models of the surface with vertical accuracy of up to 10 centimeters. LiDAR can see through objects, such as walls or trees.* we’ll discuss how later in this post

So, let’s consider an example of Airborne LiDAR to understand how the system works.

To get the height, the LiDAR system uses the speed of light, and the time it takes for the light energy to travel to the ground and back.

Now the system knows the distance between an airplane and the ground. To get the ground elevation, it takes the plane’s altitude, calculated using the GPS receiver, and then subtracts the distance between the aircraft and the ground.

There are two more things that the LiDAR system needs to consider when calculating the object’s elevation. First is the aircraft’s erratic movement due to the turbulence of the air. This movement is recorded by IMU (inertial measurement unit) and is accounted for when height values are calculated for each LiDAR return. Second is the angle of a pulse sent from the transmitter; the airborne system scans the ground from side to side to cover large areas during the flight. Some of the laser pulses travel perpendicularly to the surface or directly at “nadir,” while others leave the plane at angle or “off-nadir” (which is most of the pulses).

And last but not least: every point recorded by the system is given coordinates using the plane’s GPS receiver. And that is, basically, everything we need to know about the creation of LiDAR datasets.

Similar to an airborne system, LiDAR installed on other vehicles uses a laser scanner, a GPS, and IMU to calculate distance from one object to another object or a surface.

Two types of LiDAR are commonly used: topographic and bathymetric. Topographic systems use near-infrared light to scan land areas, while bathymetric use green water-penetrating light to map underwater terrain.

One pulse = several returns. A discrete return records individual points for the peaks in the waveform. The record can consist of between one and four returns from each laser pulse.

Now that we know the elevation measurement basics let us run through a couple of LiDAR usage examples.

Autonomous Vehicles: LiDAR works as an eye of a self-driving vehicle, constantly recording precise variable distances between a car and surroundings. This equips a vehicle with mapping and navigation capabilities.

Agriculture: LiDAR can be used to create a 3-D elevation model of a particular land, which then can be used to determine a slope or a sunlight exposure. This information will help farmers figure out which land needs more water or fertilizer, consequently saving time, money, and labor resources.

River Survey: Bathymetric LiDAR is used to measure the underwater surface. By collecting such information about the river, we can understand the depth, width, and flow of the river, which can be used to monitor the flood plains.

Modeling of the Pollution: LiDAR operates ultraviolet, visible region, or near-infrared light waves. This helps to image the particulate matter of the same size or larger than the wavelength. It can detect particles of carbon dioxide, methane, or sulfur dioxide, which then is used by the researchers to create a pollution density map of the area.

Archeology and building construction: LiDAR can detect microtopography hidden by vegetation, helping archeologists understand old buildings’ surface or maintain the record of modern buildings’ structure.

Objects Measurement: LiDAR can also be used to measure different kinds of industrial constructions. With the help of LiDAR, researchers can remotely measure and maintain the structure of roofs, bridges, and others.

Land mapping: The National Oceanic and Atmospheric Administration (NOAA) uses LiDAR to create accurate shoreline maps and digital elevation models for geographic information systems, potentially assisting emergency response missions.

Point cloud density:

https://upload.wikimedia.org/wikipedia/commons/4/4c/Point_cloud_torus.gif

The point cloud. This is the appearance of the LiDAR dataset. When downloading a dataset in .las/.laz/.ASCII formats (we will talk about them lately) what we get is many points in a 3-D space, coordinates of which are: geographical latitude, longitude, and elevation.

But before downloading datasets and creating models, it is necessary to understand the density of the points within the dataset or, as we may say, its resolution.

Point cloud density refers to the number of coordinates collected per measurement unit. The higher density, the more information we got about a particular unit. When point cloud data is processed and transformed into a 3D digital model, each coordinate acts like a pixel. So, the higher density, the higher the resolution of an image, as a result, the more accurate model for one’s particular purposes.

Point cloud density VS point cloud spacing:

Point cloud density refers to the number of points per measurement unit, and point cloud spacing refers to the distance between a point and points adjacent to it. These two terms are directly related to each other, as the lower the spacing, the more points we have in a unit, the higher the density, and vice versa.

Even though dense point clouds mean a higher resolution of an image, that does not mean that your project requires one. Attaining more points per unit can make surveying projects more costly, so before deciding on the density level, it is worth observing the project’s needs.

  • 0.5–1 pt/m²: basic surface models such as terrain models
  • 5–10 pt/m²: shapes of buildings for 3D city modeling
  • 20+ pt/ft² t: highly detailed capture of structures and surfaces

Either point cloud density or spacing you can always find in the metainformation about the dataset.

And, of course, one can retrieve the best answer to this question through an experiment.

On the Python visualization below, you can see two 3-D point cloud plots, where point density on the second plot is 10x lower than one on the first.

There are a couple of ways you can get through to the low resolution of the dataset. For example, use some interpolation algorithm or stitch two datasets together. However, it would help if you always remembered about a threshold of GPS, IMU, clock, or another part of the airborne system. Even though you can rely on matching the results within one system or company that produces datasets, there is no evidence that two random point cloud datasets will perfectly stitch together.

We are moving closer and closer to LiDAR datasets resources and their practical usage. And one of the last things you need to account for is that there are other types and formats of elevation data. And if you want to get and work with data like shown below (the practice of what will be discussed in later posts), you need to know what it is and what it is not.

With point cloud LiDAR datasets, you can use either a specific LiDAR toolset or widely-used in Machine Learning NumPy, Pandas, and similar packages. We will discuss Python ML LiDAR tools and algorithms in later posts.

Most open sources provide mapped elevation data of all kinds and formats in one place.

https://coast.noaa.gov/dataviewer/#/

Here you can see the list of available open-source elevation data kinds:

Standard elevation products are available in the following resolutions and formats:

  • 1 arc-second (30 m) DEM — GeoTIFF
  • 1 meter DEM — GeoTIFF, IMG
  • 1/3 arc-second (10 m) DEM — GeoTIFF
  • 1/9 arc-second (3 m) DEM — IMG
  • 2 arc-second (Alaska — 60 m) DEM — GeoTIFF
  • 5 meter DEM (Alaska only) — Varies
  • Contours (1:24,000 scale) — Shapefile, FileGDB

Elevation source data are available in the following formats:

  • DEM Source (OPR) — Varies
  • Ifsar Digital Surface Model (DSM) — GeoTIFF
  • Ifsar Orthorectified Radar Image (ORI) — GeoTIFF
  • Lidar Point Cloud (LPC) — LAS, LAZ (one that we are looking for)

Source: https://www.usgs.gov/faqs/what-types-elevation-datasets-are-available-what-formats-do-they-come-and-where-can-i-download?qt-news_science_products=0#qt-news_science_products

In the following posts, we will dive deeper into the topic of LiDAR datasets: where to get them, how to filter them, and how to explore LiDAR data using Python/C++ tools.

Authors: Yana Kurlyak & Andriy Kusyy at LetsData

Data Science Consultant