LiDAR Flashcards
What is ECV?
Essential Climate Variables, such as above ground biomass
What do ECV datasets support?
the UNFCC and the IPCC
What is a pro of LiDAR data?
3D
What is LiDAR data most used for in the UK?
flood modelling
What is 2.5D?
A 2D raster with each pixel containing 1 elevation value
What is a benefit of 2.5D?
It can be represented on paper without losing information
What does 3D account for the 2.5D doesn’t?
things that overlap each other
what are the three 3D data types?
point clouds, voxel space and geometric primitives
Give an example of an attribute a data point can have
properties of the object such as its colour or properties of the measurement such as laser return intensity
what is voxel space?
a volumetric pixel (3d grid wit attributes per grid cell
what do you need to consider when using voxel space?
you need to be careful about blank space
What are geometric primitives
a triangular mesh 3D dot to dot
what do geometric primitives let you do?
allows a shape to be built up, eg. spheres and cyclinders
What is a limitation of using geometric primitives?
Complex shapes can use a lot of memory
what is a useful 3D application in weather
predicting storm surges
What are the different DEMs?
Digital Terrain Model (DTM), Digital Surface Model (DSM) and Canopy Height Model (CHM)
What is the difference between a DTM and a DSM
DTM is bare ground elevation and DSM is the highest point elevation (tree tops and building roofs
Here is a list of applications
monitoring infrastructure, creating 3D models of fragile historic artefacts, ancient city footprints (Mexico jungle discovery)
What global forestry maps can be produced?
height, biomass, leaf area, structure
What is forestry data used for?
driving weather and vegetation models,
What is knowing forest structure useful for?
studying biodiversity
what can you achieve when surveying
you can collect data using GPS and total station. It’s accurate and allows context to be recorded. it can be used in forest when GPS doesn’t work
What is a limitation of surveying?
slow
photogrammetry
using views from multiple photographs
advantages of photogrammetry
Can be made using low cost cameras and rapidly collects large areas
Disadvantages of photogrammetry
requires recognisable features and you only see the top layer
which needs less signal processing. LiDAR or RADAR?
LiDAR, no need for unwrapping
Disadvantages of LiDAR
high energy requirements, limited coverage and cannot see through clouds
How does lidar work?
laser is used to emit a short pulse of light. we know the speed of light. travel time is used to measure range. GPS and interval navigation get instrument position. Attitude sensors and laser pointing are used to get direction of the laser beam
give the equation to get the distance from lidar instrument to the target
Distance = ( speed of light * time ) / 2
What are the 4 lidar platforms
Terrestrial Laser Scanning (TLS), Unmanned Aerial Vehicle (UAV), Airborne Laser Scanning (ALS), Spaceborne LIDAR
What is the accuracy resolution and coverage of TLS
~5 mm accuracy, ~1 cm resolution, ~100 m coverage
What is the accuracy resolution and coverage of UAV lidar
~5 cm accuracy, ~1-10 cm resolution, 1-10 km coverage
What is the accuracy resolution and coverage of ALS
~5-20 cm accuracy, ~10cm-1m resolution, 10-100’s km coverage
What is the accuracy resolution and coverage of Spaceborne lidar
~1 m accuracy, ~500 m resolution, global coverage
What is the most widely available lidar platform?
ALS
What is TLS good for?
small studies
What are the different ways of measuring the returning signal?
- Full-waveform – record returned light as a function of time (energy reflects back as the laser hits things)
- discrete return - Record timing of significant returns
- photon counting - use very sensitive detectors to time single photons
- phase based - use modulated output and record phase difference
which is conceptually the simplest measuring technique?
Full waveform ALS
more info on full waveform ALS and why is less common
all visible objects are recorded, large data volume and fast disk write speeds, more expensive instruments, produces volumetric data, displayed as voxels of processed to a derived product, software to handle is not quite mainstream
how does does discrete return do?
Uses signal processing to estimate the range of scattering objects.
* 1-20 ranges recorded per shot
* Point cloud produced
* Greatly reduced data needs
* All lidar software can handle
* May not record all visible surfaces, do
not know what is not recorded.
* Different instruments and survey configurations can give different results.
* Algorithms are proprietary
* Analysis can be difficult in complex
environments
full waveform vs discrete return pro and con
full waveform offers far more information than DRL, it reveals more structure>
However, at greater cost
This is why discrete is more common
Photon counting, details
- “Geiger mode” detector
- One photon per detector element
- Allows lower powered lasers (increased
coverage) - Will detect sunlight (noisy)
- Instruments currently very expensive
- sensor clicks when the conductance of the air changes
- Used by ICESat
Data formats for LiDAR data
.pts – text file
.las –binary format, a bit compressed and can be opened in arc 4 bytes per coordinate
.laz – compressed .las
Full-waveform – no dominant format yet.
.laz more info
cannot be opened in arc. Factor of ~2 saving. can be opened in cloudcompare
.las more info
, a bit compressed and can be opened in arc 4 bytes per coordinate
.pts more info
- 1 byte per significant figure (~10 bytes per coord)
how to deal with large data
process tile by tile, few Gbytes at a time, then can spread tile processing across multiple cores for “embarassingly parallel” processing
what software is used to visualise lidar data?
CloudCompare, ArcScene and Python 3D plots
What do you use to analyse lidar data?
Lastools, pylidar, ArcGIS, independent conding libraries (rLidar, lidaR, libClidar, libLas), proprietary software from manufacturer (Leica Cyclone, Riegl RiAnalyse & RiScan, QTM etc…)
1st step in processing LiDAR is…
finding where the ground is and classifying those points
what are the three classifications when identifying the ground points
ground, not ground and noise
What can you do to help identify which points are ground?
using mathematical algorithms to split up the data into a raster and connects potential ground points within threshold you choose - the algorithm does this until it cannot add any mire points
Criteria for ground points
- Ground will be at the bottom
- Within a given small area, ground points are likely to be at the same elevation
- The ground is unlikely to have sudden spikes or deep holes
- Changes in elevation are likely to be smooth
LASTools lasground_new.exe function used in the practical
Three most important variables
step, spike and offset
step
How large an area are statistics calculated over. setting the size of the raster, AKA ‘patch’
- If short, it will find sharply varying surfaces, but will pick up building roofs as ground.
-spike/-spike_down:
How large an instantaneous change can still be classed as ground
Smaller than a wall but larger than the depth of the plough in the case of the field
Offset
How much variation do we expect from the line of best fit?
- If data has any ground roughness (eg ruts in ploughed fields) or is noisy, this number should be increased to contain the variation in ground elevation.
Offset must always be bigger than the noise
how big is the uncertainty in object location given the range uncertainty?
~ 20 cm
what issues to building cause
- Large areas with no ground returns
- If patch size (step) is less than building size, lowest point will be on the roof.
- Need to adjust patch size accordingly.
- Can use sudden change from ground to
wall identify building.
What is the rule of thumb when covering an area with large buildings?
Rule of thumb is the step needs to be at least half the size of the largest building in your area
Issues with vegetation
Trees can block view of ground – requires
longer patch size than bare-Earth.
* Forests can have complex topography (too steep to farm) – requires shorter patch size.
* Where does the ground end and the plants begin?
Point density is important to determine where the ground is
issues with complex topography
Rapidly changing steep terrain
* Real ground can have sharp changes
* Requires a shorter patch size
example - quarries
increase spike to pick up these real changes
Issues with ALS density
Point density is key
* The more points, the greater the chance of all objects being correctly measured.
* Do we see the real ground?
4 point/ m^2 needed to see the ground well
What can be used to estimate population in countries without census
Building volume
Datums
Our elevation is measured relative to a “vertical datum”.
This is most commonly one of:
* An ellipsoid (assumes a regular shape for the Earth)
* Mean sea level through the local gravity field (“geoid”)
MAKE SURE YOU KNOW WHICH ONE YOU ARE USING, OTHERWISE ALLIGN THEM IF COMPARING
Advances in ground finding algorithms
In the last 20 years they have become more sophisticated
Classifying surfaces
Use texture metrics – standard deviation of points about the mean elevation.
Buildings will be smooth, have lower stdev.
Vegetation will be rough, have higher standard deviation.
give an example of a spaceborne LiDAR looking at veg
Global Ecosystem Dynamics Investigation Lidar (GEDI)
First lidar mission designed for forest measurement.
* $94,000,000 mission
give an example of a spaceborne LiDAR looking at ice
ICESat - 2
Optimised for ice measurement
* Photon-counting, greater coverage but noisier data than GEDI
* $1 billion mission
give an example of a spaceborne LiDAR looking at clouds
CALIPSO
Optimised for cloud measurement
* Full-waveform
* Dual wavelength and dual polarisation
* 2006 - 2023
used for forecasting
give an example of a spaceborne LiDAR looking at wind speed
Aeolus
* Uses high-energy UV laser pulses
* Data is used to improve weather forecasting
summary, what 4 main things can lidar measure wind speed
- 3D structure
- Surface elevation
- Height
- Wind speed (with Doppler)
measures doppler shift off the gas
also used in forecasting
What are level 2 products?
things that lidar directly measure
What are level 4 products
Things you are modelling, inferred through statistical modelling - eg. biomass
Much lidar is now freely available, where?
- England & Wales environment agency
- Lidar Scotland - https://remotesensingdata.gov.scot
- Open Topography
what country gets renewed coverage every 6 years?
Finland
what is a limitation of the lidar industry evolving?
power intensity
what is evolving
detector efficiency
Which gas blocks more? CO2 or methane?
methane