Over fifty percent of the people on earth are currently living in urban environments, according to UN data. This flock to the city is expected to continue, with a projected urbanization of 68% in 2050. Governments all over the world prepare for this urban growth. The European Commission, for example, has stated the will to reduce urban sprawl. What we have known for a long time now in Groningen, namely that more dense cities are more liveable cities, appears to be widely recognized. Influential urbanists like Richard Florida call for the need for densification, but how does such densification change how we perceive our urban environment?
(very) short history of mapping
One way of perceiving our environment is through the use of maps, and this article aims to provide insight into mapping complex urban environments with a focus on indoor spaces. The Netherlands has a history of hundreds of years of high-quality mapping (do you know www.topotijdreis.nl? If not, check it out!). Because of this long history with mapping, the Netherlands has grown particularly good at it. Nowadays, the Dutch Kadaster is even actively helping other countries to develop similar information products. Furthermore, services like Google Maps and OpenStreetMaps are used by a large public on a daily basis, and there is a big chance you have used or will use them even today. They enable users to plan trips, navigate and explore their surroundings easily and on the go. However, all of these maps are preliminary based on data that is outside, documenting for example roads, buildings, and parcels. Indoor environments, where most of us spend most of our time, remain black (or red, in the figure below) boxes.
Densification leads to high rise
One consequence of the wish to densify our cities is an increasingly complex urban fabric. Six out of ten of the largest buildings of Groningen have been built in the last ten years, while the housing situation urges more high-rise to be realized in the upcoming years. It is not hard to imagine that high-rise buildings are more complex compared to low-rise buildings, as the buildings can serve multiple use cases, harbor more people, and offer multiple entrances and routes to indoor destinations. As people are now used to location services in the outside environment, demand grows for similar services in indoor spaces. This entails, but is not limited to, indoor navigation services and indoor asset positioning services (like finding a book in the library: https://www.rug.nl/library/news/190923-wayfinding-app. Although the development of these services is expected to hold benefits in terms of efficiency, there are also more urgent use cases to be thought of, like emergency management.
If disaster strikes, be it a fire, an earthquake, or a man-made public shooting, a whole series of activities will be put into action. First responders like police, fire, and medical departments will be on their way to the incident as soon as possible. Due to high-rise buildings, their operational environment is getting more complex as well. Taller buildings complicate things like using a fire hose, as these hoses can only be of a certain length. These environments do also require first responders to go into the buildings to help people, which is a severe security risk. If little information about the building is available, the operation is even more dangerous. This is the case for single high-rise buildings like the ones in Groningen, but it gets even worse for high-rise districts like ‘Kop van Zuid’ in Rotterdam.
At Kop van Zuid, several skyscrapers are situated on a small piece of land, only connected to the mainland by the Erasmus bridge and a walking bridge (figure 2). This means that there are many people living and working in a very dense area, while the entrance/exit routes are all the same. An evacuation of the plot would, therefore, be very complex. Indoor maps could increase situational awareness of first responder organizations, which would smoothen a process of an evacuation. You may ask yourself, if these maps are of such good help, why are they not readily available?
Indoor mapping process
There are several reasons for the unavailability of indoor maps. Sometimes, floor plans are available. Floor plans give a clear overview of the layout of a building, although there is often a lot of detail missing, like fire hose connections and staircase connections. Safety region operators are dependent on these details, like geographical information about windows and other spatial information. This includes height information, which is why the data is preferably available in 3D. Building Information Models (BIM) offer these details. BIM models are 3D digital information sources in which not only the geometry of objects like walls and doors has been specified, but also semantic information like the material that has been used for the wall or the exact location of electricity fuses. These properties are very valuable for first responders, but not all buildings have a BIM model. Furthermore, BIM models are used to plan and maintain buildings, but this does not mean they are always up-to-date. Buildings are dynamic structures, which means that they are changing all the time based on their use. This is difficult for first responders, as they depend on the accuracy and reliability of the data. Frequent updates are therefore required.
There are several ways of creating indoor information. Of course, one could measure and draw the geometries within a building by hand in a digital environment. To this geometry, attributes like height and material can be added. As you will understand, this is a time-consuming task and therefore not very suitable for large-scale mapping of buildings. Furthermore, frequent updates are required to keep the model up to date, which is again a cumbersome process.
As an alternative to this manual process, many companies are researching the possibility of automatic processing of indoor maps. To indicate how such a process works, this article will describe a general process of automatically mapping a building. Of course, individual methods might derive from this general process.
First, a scan is made of the building in question. This is often a 3D scan with a so-called ‘point cloud’ as a result. This point cloud is the raw input for the 3D model and gives a very ‘real-world’ representation of the building. A point cloud is quite literary a dense ‘cloud’ of points, each with an x-, y-, and z-coordinate to depict their position in 3D-space.
They are often acquired by the use of LiDAR, a laser-based distance measuring system. Originally, these systems were always placed on a tripod on a fixed location, from which a space like a room can be measured in millimetre precision. The scanning system rotates in a 360-degree fashion, rapidly firing one laser beam for every angle of the scanning system. If a laser beam hits a surface of some kind, the beam bounces off, partly back to the scanner. The time of return (or non-return) of each laser beam is measured, from which distance from the scanner to the surface can be calculated. Combined with the angle of measurement, a point (x, y, z coordinate) can be placed in 3D space. As the scanner takes many, many measurements, a ‘cloud’ of points results. After finishing the scan of the room, the tripod, or ‘terrestrial laser scanner’, has to be moved to measure other parts of the building. Although the resolution of such a scan is very high, you can imagine that moving the tripod each time is not very efficient. To take care of this problem, Mobile Laser Scanners (MLS) have been developed. These scanners use Simultaneous Localization And Mapping (SLAM) algorithms to be able to scan while moving. These scanners can, for example, be taken by hand while walking through a building, making the scanning process a lot more flexible. The resolution of the scan decreases, but, on average, centimetre accuracy is still possible.
Next to LiDAR scans, photogrammetry can be used to make point clouds as well. This is a method based on calculating distance values from combinations of images taken by ‘regular’ cameras, like the one on your smartphone. For example, the overlap of the photos can be combined with orientation data to add depth to the pixels, again resulting in a point cloud. Alternatively, instead of a point cloud, a mesh can be created, which is a connected (structured) point cloud.
Although the processing of the scan into a 3D model is very interesting, it would take a lot of space and technical details to explain it. The most important notion of this is there are two ways to segment the point cloud into recognizable objects. One way is to identify geometric forms. A wall is almost always a vertical pane, while a spiral staircase is better identified by fitting it to a cylinder. Algorithms like ‘Hough’s-algorithm’ or ‘RANSAC’ can be used for this. As a second way, deep learning can be used for identifying features within the point cloud. A digital neural network is trained by showing it images or point representations of objects (like chairs), and if this training phase is done well, the process will be able to recognize similar objects ‘on the fly’.
Visualisation and results
Finally, the result can be visualized in a 3D environment. As the result is positioned in 3D space, the result can be viewed from different angles, and distances can be calculated within the environment. This is much more interactive and therefore of use to users like first responders, who can now calculate the length of the hose needed for extinguishing a fire within the building. Of course, a nice addition is that you will never be lost again!
The original article was written two years ago. Nowadays, I work as a Geo-ICT Business Consultant at CGI Netherlands. As you have probably noticed within the curricula at the faculty of spatial sciences, the digital era is rapidly closing in on the fields of spatial planning and geography. Are you interested in digital transformation? Feel free to reach out and send me a message at Bart-Peter.Smit@CGI.com.
This article was first published in the Girugten Lustrum Edition (Year 50 of Girugten – issue 02 – May 2021) under the title: ‘Mapping the indoor urban fabric: the next challenge’.