A Country-Wide 3D Model of Belgium at the Price of a 3D City Model!
Author: Andrei Efimov
This blogpost is part of a series of blogposts written by Ordina’s GEO-ICT team. Missed the first part? You can read it here.
The Silver Bullet, Part 3
If you are still here, it means that our first two blogs raised some interest or at least some questions about MY3D. So, your logical next question might be… how much does a product like this cost?
Well, it requires a lot of investment for Ordina, but we are thrilled to announce that we can offer the entire country-wide model at a comparable price of a custom 3D city model. Yeah, I know, Belgium is not so big :D
We see on the market, that for already created and available for sale city models, prices vary between 100 and 1000 euro per square kilometer, depending on the quality of the model. Custom-made models, created especially, let us say, for a local authority, will cost tens or even hundreds of thousands of euros for highly detailed models. Prices are higher if a flight is required to gather new data and if the models become the sole property of the customer. It can drop if existing data is used or no manual editing of the models is performed.
You must already be thinking that something is off… you cannot offer the entire country for some tens of thousands. Yes, we can, our business model allows it. We can offer it because:
· We use open lidar data as the raw input for processing our models
· We extract our models using automatic procedures, with very limited manual editing.
· We make the data available to multiple customers, each with their specific needs.
So, we will return our investment in time, by attracting many customers from different sectors.
And this is not the only good news…. If customers do not need all the information linked to our models or do not need all the display and analysis capabilities of our platform, prices can also be flexible. If you only need data for a specific location, like a commune… the price will be only a fraction.
MY3D data can be bought, or offered as a service, and used only as long as you need.
You might then wonder that maybe the quality of our 3D models must be very bad at this price. The answer is no, it is quite good. How good?
I will answer this with a question. How accurate should a 3D model be?
It is quite simple, it depends on the customer needs, on the business case. An architectural or construction company will need very accurate models, survey grade. A model used for visualization or presentation purposes by a local authority can be much less accurate, other uses are more focused on the data attached to the models than to the geometry of the models.
Today we could talk about a few cm accuracies, using for example terrestrial laser scanning or even mobile mapping or UAVs, but prices for such products restrict them to small areas. On the other hand, models too inaccurate or missing important details are not usable. For example, you can find LOD 1 models even for free. But an LOD 1 model is just a cube, it gives no information about the building roofs. If you, for example, are interested in building renovation or energy efficiency, those models are useless.
On our first blog we mentioned that we see use cases for many sectors, but especially for Energy, Utilities, Finance, Immo, Retail or Public Authorities. Therefore, the quality of our models is tailored for the business cases of such customers. In most cases, a customer would need LOD2 models with mapping grade accuracy.
In fact, we like to call our models LOD2.2 models, due to the level of details we model on our roofs, which goes up to 4sqm. Even smaller details on the roofs will be visible in the models (Science direct, Filip Bjleki)
Ok, the details are there, but what about spatial accuracy? As a general rule, our models cannot be more accurate than the raw data used to create them. The model is based on many data sources. However, for the spatial accuracy there are two that are the most important: GRB vector data and DHMV II lidar data. These two, plus similar ones for Brussels and Wallonia have a good absolute spatial accuracy, usually between 5-20 cm.
Based on the accuracies of these raw datasets and considering the errors added during feature extraction, we can state that our models have the following spatial accuracy:
Planimetric accuracy (x,y): 35 (cm)
Altimetric accuracy (z): 30 (cm)
*calculated based on Oude Elberink and Vosselman (2009) methodology ; **sigma1 confidence level
According to ASPRS, this puts our models in the mapping (planning) class I accuracy category, the highest non-engineering/surveying class.
This is more than enough for the types of business cases we foresee for our models. For example, for a renovation estimation, knowing your roof height with a 30 cm accuracy will do the trick nicely.