Mapping Major Cities with LiDAR: The Future is Here

Knowledge 20200127 2

Drone-mounted and airborne LiDAR are changing the way architects, city planners and construction surveyors map the world’s major cities.

Atteyeh Natanzi, PhD at and Iman Zolanvari PhD were post-doctoral researchers at University College Dublin’s School of Civil Engineering when they worked on a major urban LiDAR survey project. In a recently-published GIM International article, they revealed the details of their LiDAR coverage of Dublin over two studies done in 2009 and 2015. (Consult the results of the 2015 Aerial Laser and Photogrammetry Survey of Dublin City Collection Record here.) This interview has been edited for clarity and length.

Laser scanning point cloud of Dublin, Ireland’s capital city (courtesy GIM International)

YELLOWSCAN: You argue that because more of the world’s cities are becoming megacities, LiDAR (and to a certain degree photogrammetry) is already the wave of the future with respect to sustainable growth planning. Why is the use of 3D mapping so essential?

It’s very important to have accurate geometrical data in a city where there is a lot of automation, from self-driving vehicle navigation to optimizing traffic flow, to automated delivery systems using drones and robotics. Having accurate geometry is crucial for smart cities that are or will be integrating lots of sensors, so we need accurate 3D models of the city. It’s also important to keep building LiDAR footprints updated. When you have a lot of construction in developing megacities in say, China and India, the Middle East and even in western Europe, we need the latest LiDAR data.

YELLOWSCAN: You both worked on the 2015 collection of Dublin data via helicopter-mounted LiDAR: How long did it take, from planning the flights to producing the data?

This was planned from 2013 by Professor Debra Laefer in University College Dublin (UCD). We joined her group to start our PhDs. She was awarded funding from the EU Research Council to capture that LiDAR data working in the urban modelling group of UCD. There was a lot of prep and calculations; planning the flight routes and elevation and all of these tech calculations. It was like a chessboard in that there were two series of flight strips, like a chessboard from one corner diagonal to the other corner, horizontally and perpendicularly, around 20 strips in two directions. It was interesting that our colleagues opted to capture the data during the winter in order to minimize shadows of buildings and leaves on the trees for minimal vegetation. It required a half-day to capture performed by an outsourced company and was delivered a few weeks later. The raw data set is publicly available at NYU open data repository.

YELLOWSCAN: What were some of the uses of the Dublin survey, apart from the summary article that you co-authored in GIM-International magazine?

Several papers were published including in the ISPRS Journal of Photogrammetry and Remote Sensing that explored the further usage of the data set. Prof Debra Laefer also published more than 10 papers on that LiDAR data set.

[Iman Zolanvari] After I got my PhD I joined Trinity College Dublin (TCD), and I proposed an idea of labelling the LiDAR raw data set. The raw point cloud itself was unstructured and only had geometry of points X Y Z in the 3D space, as well as the intensity of the laser beam that was returned to the scanner. The contracted company itself did a primitive classification of the vegetation and buildings, but obviously it wasn’t enough. What I proposed to do, and accomplished with the help of 21 people, was manually process during over 2,500 hours big chunks of data sets in three coarse categories, which are BUILDINGS, GROUND and VEGETATION points. The BUILDINGS category is more detailed and labelled into façades, roofs, windows and doors; GROUND is divided into three categories: pedestrian, street and grass; and VEGETATION included all trees and bushes. Whatever remains are UNDEFINED points (e.g. bin, bench, cars, etc).

It’s a new annotated urban data set for such density and coverage area which is designed for several applications. The most important of them may be employing machine learning techniques. If a car wants to drive in a city or to do forest monitoring, or even for urban planning and emergency management and all of these applications, we need to understand the 3D scene. Even in business marketing, LiDAR has a lot of applications, for instance, if a company wants to install double-glazed windows, they’ll want to approach the buildings that have the most number of single layer windows [and get an accurate map of it]… therefore we need to know among those lidar data sets which are exactly those structural elements.

So with the annotated Dublin LiDAR data set, we manually labelled [the point clouds] in hierarchical order. People with different needs, with different applications, can approach and extract all of the structural elements (for example windows or all the roofs) of the Dublin City Center, and then they can use the information to train a Neural Network for labelling and extracting the urban features.

YELLOWSCAN: When surveyors use LiDAR for archaeological purposes, they often make surprising discoveries. Were there any surprises or discoveries about Dublin as a result of your LiDAR and photogrammetry survey?

The surprise was first of all the high density of the point clouds, around 300 points per square meter, which after five years is still one of the densest data sets available for a large urban area. Also, in the Liffey River and surrounding elevation you can clearly see the average elevation and height of the area. In that first image of the GIM article, you can see the colorized format of a data set where we added the color regarding its elevation of the points. You can clearly see that the south of Dublin has on average a higher elevation than the surrounding of the river. So imagine god forbid you have a flood from the Irish Sea towards the dock? You can see clearly the lower areas are the places that most likely need flood management and could be at a higher insurance risk, and for urban planning as well. Another thing I really like when you look at the data is that when a laser hits the water, it receives two returns from the laser beam: one is from the surface of the water and the other is from the bed of the river. It is also interesting for mapping and calculating the volume of the canal for discharging rate.

YELLOWSCAN: If you were able to conduct a similar survey with drone-mounted LiDAR today, would it be faster and more cost-effective?

Regulations restrict the use of drones in the cities. Finding people who are licensed isn’t easy or cheap. There is one guy in our group who develops and uses drones but mostly for photogrammetric purposes. Generally speaking, the technology is advancing, LiDAR scanners are becoming smaller and cheaper…For example, at Maynooth University there is a GIS group that has a drone-mounted lidar and they can capture different areas, it’s good technology for testing and capturing a targeted area.

YELLOWSCAN: Can you tell me a little about your current LiDAR research? What’s coming up?

[Iman Zolanvari] I’m working at Ambisense LTD, a start-up company developing IoT sensor. They are developing cheap, practical solutions for capturing data in smart cities. For example, using small scale radar devices that are pretty cheap and can capture, process and transfer 3D point clouds in real-time.

[Atteyeh Natanzi] I’m a post-doctoral researcher, now working in an EcoStructure Project that has large groups in five universities across Ireland and Wales. We are trying to find ecologically-friendly marine structures. We have used LiDAR for mapping seashore rocks and structures.  We have LiDAR point clouds from both sides of the Irish sea, we use that data in this case for modelling and monitoring marine life. We will use the LiDAR data and make a 3D map of the Irish sea and 3D printing moulds for sea defense concrete tiles and even printing the rock shape cement.

NB: Author Jordan Robert.

Discover our UAV LiDAR Solutions