SDSW23: Denmark in 3D: Scaling 3D Data Production to the National Scale

tl;dr

This presentation about the Denmark in 3D project is a collaboration between 1Spatial and SDFI. The project aims to create a 3D model of Denmark for various applications, including simulating sea level rise and climate change impacts.

The speakers discuss challenges such as data quality, sustainability, and political implications of implementing 3D data. They also describe the technical aspects of creating and validating the 3D buildings, as well as the incentive for municipalities to upload their own building data.

The project is seen as a way to make 3D data more accessible and usable in a business-as-usual process.

Transcript

Phill Ridley: So, we're here today to talk about a really, really innovative projects that 1Spatial and SDFI have been collaborating on, which is the Denmark in 3D project. And there's an increasing driver for 3D data in the market in geospatial. And one of the primary drivers of that is digital twins.

So we're seeing digital twins are coming up and up over and over again in the conferences that we all attend. And when I think of a digital twin, the first thing that comes into my head whenever I hear about digital twin, digital twin being presented, is my grandma. So there she is. This is her at my wedding she was a few shandies down, I must admit, when this picture was taken. But a couple of months ago I had my grandma over for dinner. She's 94 years old, and at the end of dinner she insisted on washing up all the dishes because she likes to help out. And I told her, Grandma, you don't need to worry about washing up the dishes. We've got a machine that does it for you. A dishwasher. As she could not, she could not believe what I was saying. She said, Phill, you've made it in life. There's nothing left for you to achieve more now. You've got a machine that washes your dishes. And I assure you, at the young age of 31, I hope to achieve a bit more in life than owning a dishwasher. But she was so taken aback by this, and it got me thinking when my grandma was growing up, we had a family pub up in the Lake District, and if you had a car, if you had a washing machine back then you were doing really well in life. And these are all things that we take for granted. And back from when my grandma was my age up in the Lake District, technology has come on leaps and bounds. We've got smartphones, all sorts of awesome stuff. But my grandma has got all this technology that's made available to her, but she hasn't got any wifi, she doesn't have to use the Internet. She can barely use a smartphone if I lend one to her. Her infrastructure, her underlying foundations to support the technology innovations have not kept up. And that's why when I think of digital twins, I think of I think of my grandma.

There's so many cool technology innovations that are taking place, particularly with digital twins and climate change. So what we do is we use digital twins to build a replica of this world, of a city of a town, and then we can use that to simulate things like sea level rise, evacuation procedures for the emergency services. And we start to use that to create a digital version of this world to see what the future might be like if we throw certain scenarios at it and with the instance of climate change also, we can start to work on how we going to mitigate the impacts of climate change. Where's the best place to be putting renewables? If you're putting solar panels on a roof you want to make sure it's pointing towards the sun and you go to all these sorts of conferences all over the place and you see all sorts of really cool innovations. You can see there a city that's been submerged by the sea. Augmented reality.

This one at the bottom. I was at a meeting three weeks ago and we got presented with a new product that's coming out called The Goteverse, right? So we use a digital twin at the moment. It's a plan for the future of the real world. Someone has said, Well, the Metaverse is becoming a like a digital version of the universe. Therefore we need a digital twin of the Metaverse to predict how the Metaverse is going to be behaving in 20 years time, and I feel like Leonardo DiCaprio in that film Inception, because we're getting these layers upon layers and layers upon layers. And you go to these conferences and we've all had it in our careers. I had it in my last company before I joined 1Spatial. Your CEO goes to a conference, they see the Goteverse, and they go that's awesome. I really want that. Can you get that implemented for me? You go yeah no worries at all, and you get your business case together, you get your plan and then you go to your CFO and they go “No chance. Here's a budget for a Ford Fiesta. Give it your best shot.” So what we're have to try and do and because of this, because of the wants of senior management compared to what is actually realistically possible in the world, the national mapping agencies are realising there's now a need to start to be supplying this 3D data to make these sorts of innovations possible. And we're starting to see this really take off across all of Europe.

And the challenge is how can we start to get 3D data available in a business-as-usual process? So not just one-off innovation projects, but how can we get it into a business as usual process? And it comes along with the same sort of things that we always see with geospatial data. The first one is one of the constant problems - data quality. We see it all the time, but when you're going from 2D up to 3D, the problems that you've got just go through the roof. So if we're producing 3D data for other organisations to be using at big-scale city-wide nation-wide, we have to have really robust data quality processes in place. The next one - funding, always fun. I’ve covered that before.

The next one that's really interesting now when you get into the 3D space is data sustainability. So we always talk about data quality, data, currentness, but also data sustainability. When you're producing these 3D data models, you can't you can fly over the country, you can capture it, you can build a 3D model and you go “Brilliant, we've done it.” But one or two days later, that model is already going to be out of date because you see it all the time, particularly in specific cities and urban areas, buildings are being knocked down. They're being extended, they’re being split into different parts. So we want to find a way that we can make sure that our dataset is sustainable. That your CFO’s you're struggling to get the check out of them to write that initial database, let alone doing it time and time again.

So how can we be finding ways to make sure that our national datasets is sustainable with change only updates? And there's a really good rules engine for doing that on exhibition upstairs. So I'd recommend you go and have a look. And the last one that we need to think about is politics. Whenever you are trying to make a decision, you know that the more people that get involved, the harder it's going to be to come to a conclusion. Every leader of an organisation is happy to collaborate with another organisation as long as all the other organisations do it their way. So as soon as you start to have 3D models and you've got people submitting data in and then you've got people having impacts on the uses and use cases of that data, it gets really politically, it gets really politically challenging.

So that's just the initial introduction. 3D data has always been that thing down the line that may or may not be coming. It’s starting to come now, particularly in the European countries, it’s really starting to take off and what's really exciting about this project is that we're looking at how can you take 3D data from those cool things that you see at the conferences into actual business-as-usual processes. And Peter and his team are the clever ones who've been actually working out how are you going to do that? So I'm going to hand over to Peter now. He's going to take you through some of the technical sides of the project and how we've been getting it all implemented.

Peter West-Nielson: So now to the boring stuff. I'd tell you a little about how it is in Denmark. We have like we have no national datasets, 3D dataset. We have like 98 municipalities, like 20 of them are big enough to solve sometimes make a 3D dataset, but they haven't got funding for updating the dataset and they don't, they only make part of the municipalities in a 3D dataset. So in SDFI we have a service model which we update each one fifth a year and we have topographic datasets. And so for this project, we want to create the 3D buildings using existing data to bring the cost down, to make a national model. Of course, that doesn't really generate the perfect buildings. We know that a lot of the buildings we will generate is not perfect. So around 90% of the buildings will be okay and there will be some 10% that is strange looking because of the point model, the elevation model is so somehow strange. Or the mechanisms we use in the building reconstruction, that's a German company that makes a 3D buildings, can't make a good looking building 3D buildings, for instance, churches and cathedrals and architectural buildings look strange we can't really generate a perfect 3D building, but we accept that because this 3D building is not pre-made. It's mainly for good looks. It's created to be able to generate other calculate other information from it.

Yeah, so if we look at the overall architecture, we will have a 3D production system. I will show you that in a moment. Then we have our data bank, our data warehouse where we have all kinds of data stored and the production system takes data from our data bank and our data warehouse, generate the buildings and put them back into the data warehouse. And then we have our data supply where we can then distribute those 3D data. And then we want to have a system where municipalities can say, well, we actually want to have a better version of the church or the municipality or the yeah, whatever the museum.

So we give them a system where they can then upload a better version of that building if they think that's important for them. But we will not do it at the first take. And then of course we have found out that it's very important for us to have actual specifications that defines “What is the quality we want to have, what is the detail of the 3D buildings?” Because, everybody thinks different about the buildings, the detail and we have had a lot of City GML files which have been uploaded to us and they look different each time.

So we found out that we need to have a specific specification defining precisely how the 3D buildings should look like to be able to look at. Yeah. And what is the specification says something about the system. how do you upload the building, what is the requirements for the data? When you upload some data, you both upload data about the buildings, but you also have to upload information about missing data. Who is uploading? What have they created the building from? And all this stuff. And then we have some requirements about the geometry, the topology and we have defined how do you actually update the building so you can just send us the building and if it has an ID that we already have in the system, we will upload the building, but you can also give us the information about it, should it be deleted, updated or is it a new building? So we give the option to tell us about that and we then have those information to upload the update database.

Yeah. And then we have this, upload page where you can then upload the data. And what you upload is of course the City GML file. We defined that it should be City GML because that's sort of the standards in the community. And then we have made them a metadata delivery file. It's not very big, but that tells you something about who is uploading and what is the email. And so we can send information back if, if somehow we can't upload the data because there are some errors in the file.

Yeah. So if we look at the production system, we will have a first, we will have a go where we just create buildings from the entire country, from existing data. We have this automatic building creation system and then we have an evaluation module where we compare the 3D building with the point cloud and try to give a value. How good is that building? What's the reality? So we can sort of, if a municipality wants to have better buildings, they can sort of pick out the we want to have all the, all the bad buildings they can send to, let's say private company to have them create a better version of the building, which then they can upload afterwards.

Once we have created the building, it could go into the building validation system. And if the building and the system can be updated, then we can update the database and keep all the versions of the building so we don't delete anything. We keep all the different versions of the buildings. So we actually do this with the 3D pillars, as we do with every other datasets which we have in the house. We keep a version of all the data. And if we look at the building validation and database update.

So the first thing that happens is we do a simple schema validation, which we use FME for and if that goes okay and if the major data system is okay, then it goes further to the geometry and topology validation which we use the 1Integrate tool for, to upload to validate whether the topology is okay. And the geometry of the building is okay. And then we do the database synchronisation where we, yeah, we make a version history, the historic version of the old building and putting the new one in. And if anything goes wrong in this validation, we will send a status report to whoever has uploaded the building and try to explain why can't you upload the building or the buildings. And then what's the reason why it went wrong. Yeah. And the new building and it will, compared with the old building and will be replaced and we still have the old version of the building.

Originally the idea was that we want to keep the ideas of identification of all the buildings, and we wanted to also keep the identification of all the surfaces. So for instance, if you have a, let's see, an IOT device that is placed somewhere on the wall, if you have updated the building, it's only three of the four walls that is updated. Then we can still have the ID identification on the last surface. But we are not that ambitious at this time.

And this is some of the things we want to search for. We want to see the requirements of how the plane of the building and yeah, the position accuracy, point density we want to have yeah, we don't want to have too small buildings. And so that's some of the stuff we want to use to validate the buildings for the some yeah, some more. At the moment it’s not very extensive, but I could easily see that the more the system is being used, the more we figure out that we need actually some to have some more rules, but then we have to to change those specifications.

And lastly, we have the 3D data model, but we have decided to use to create our own model, because we want to have the 3D data look exactly as we do every other data set. So yeah, a building has this information about how good do we see it is and it has every building has to have a reason to what, how did we upload the data. What, who delivered those buildings. and then of course we have references to all the registers so we can. The 3D data model doesn't really contain that much information, but because we have a relation to other datasets, we can combine the attributes from those other datasets.

Yeah. So that's sort of the idea we're working with at the moment and testing.

Phil Ridley: This is the beginning, the beginnings of a bigger project. What we're doing at the moment is, as Peter said, we're working through each of these steps. We're scaling it up bit by bit, identifying those extra rules and things that needs to be built into the data quality process and the data governance process. Because when you do go into 3D, I'm sure, as Peter says, it is another whole level of complexity. So please do come and have a chat to us. If you've got any questions or if we've got questions now.

John Harsthorn: Yeah do we have any, so thank you very much Peter thank you.

Member of the public: Great presentation. Thank you so much. My name's Sam. I'm just I get really confused between two terms. One is 3D model and the other one is digital twin. I wonder if you could tell me how to get from one to the other. What's a model and what's not a model? Thank you.

Phil Ridley: I can tell you the joke answer. I can tell you the actual answer. So the joke answer is you say, digital twin to get people to come to your marketing event because it's a buzzword, but the actual answer is the 3D the 3D model is the skeleton that sits underneath. So a digital twin is very much about where is data coming to and from. So it's all about the data processes. So for example, you're reading data in from whenever you're walking around and you're collecting data, you're taking pictures on your smartphone that would be geotagged, that would be then going into what is called a digital twin, which is collecting data, a digital version of this world. With a 3D model, it’s a skeleton that underpins that because you need to then if, for example, if you're taking a picture at the top of the skyscraper on a rooftop bar, you need to have something that you’re then geotagging and referencing that against. So that's where the 3D model comes in, it's that skeleton on which the flesh of a digital twin can go. Does answer your question.

Member of the public: Yeah, that's great. Thanks very much.

John Harsthorn: Okay. We have learnt about skyscrapers there and rooftop bars, which we don't have in North Yorkshire. Another question?

Member of the public: Sorry if I'm a bit confused by this, but what's the incentive for people to upload their buildings?

Peter West-Nielson: To have a better version of the building. I mean, if that's important for their use case that the church looks better than what's in the national data set, then they have the possibility to upload it. That's then sends it I mean if they want to use the 3D dataset for presentation or communication, stuff like that, perhaps they want to have a better version of, let's say the church, okay. And we give them then the possibilities to do that. Right.

Phil Ridley: And sorry, just to build on that as well. What we're doing here is we're creating a national module which is accessible to everyone. So the different municipalities will have different budgets, different amounts of money. So you might have one municipality that's got lots of significant government buildings of importance. For those, you'll need a really detailed model for counter-terrorism prevention and things like that. But if you've got another municipality that doesn't have the budgets to do that but can see the value of a 3D model, they can just use the national one.

John Harsthorn: Right? Thank you very much, Peter and Phill.