CGarchitect interviewed Vertex Modelling, a London based company building the most accurate and detailed 3D Model of London. The model stretches across variety of urban landscapes from the new skyscrapers in the City of London and Tower Bridge to residential houses in the boroughs of Chelsea and Knightsbridge along with a variety of London's well known and iconic Landmarks. The Interview was conducted with Vertex Modelling manager Paul Quaife. (Vertex Modeling Profile on CGarchitect)
After completing a degree in Computer Aided Product Design I took a year out to travel. When I returned I fell into a role at the Ordnance Survey on temporary basis within the Photogrammetry department which I loved. After 18 months I started in a permanent role within the Cartography department. It was here that my love of 3D design and mapping/GIS merged and I began to experiment with OS data to add the 3D element. The obvious link was photogrammetry, at the same point there was an opening at GIA in London to create 3D models from aerial imagery, I applied and the rest as they say is history. The rest of the team have either history in Photogrammetry, Geography or 3D modeling. We all share an interest in Architecture and photography, and one member of the team landed a job through Twitter!
CGA: How did you get the idea to build the London model?
Quaife: We started building the London model back in 2007. At that time, construction was booming and there was a massive demand for small scale models of development sites. Our largest customer was a London building consultancy GIA who required an accurate 3D model for Rights of Light, Rights of View and Shadow/Glare analysis. That meant that we were forced to tailor our model to their strict requirements of high detail, high and consistent accuracy and high usability - by usability I mean the ability for them to prepare the model for analysis as well as the ability to print our models on our in house 3D printer. But these constraints also meant that the quality of capture and post-processing was far superior to everything on the market, realising that we had spotted a gap in this niche market we set out to create a wide area model of London. The first step involved connecting the smaller sites we were working on and then expanding the model on a project-by-project basis.
CGA: So what is the process behind building this model?
Quaife: A small but state-of-the-art aerial-based photogrammetry flow line is at the heart of what we do. Although the costs associated with acquiring the hardware, software and source imagery needed for the set-up was considerable, there was no other way to meet all of the requirements set out by GIA. We were lucky that at the same time we were setting up our operation new digital ultra-high resolution aerial imagery with GSD of 4cm became available, which made it possible for us to be accurate to within few centimetres on all axis. Combining this imagery with a large number of ground survey points allowed us to create a central data set from which we are able to derive our 3D models from.
CGA: Aerial-based photogrammetry, Aerial Triangulation, 4 cm GSD - this all sounds very alien to CG Artists. Can you explain little bit more?
Quaife: Sure. Stereo Photogrammetry is the practice of determining the geometric properties of objects from photographic images - in our case, the geometric properties of terrain, buildings and manmade features. Basically, an airborne vehicle, typically a plane, is equipped with a calibrated high-resolution camera and flies over a target area taking continuous photographs. Each of these photographs has metadata associated with it captured by an internal measurement unit (GPS coordinates + pitch, roll and tilt of the plane). We use this data to position the photographs in relation to their real-world location. Because these photographs include a 40% overlap, we are able to tie common points from adjoining photographs together - this process, together with adding control points and ground-based GPS is called Aerial Triangulation. We then view these stereo pairs using 3D stereoscopic displays (one photograph into one eye, second in the other eye) and trace the features as we see them in this 3D environment. To do this accurately you need photographs of ultra-high resolution - this is where the term GSD (Ground Sample Distance) comes to play a role. You can view on Google Maps the industry standard aerial imagery for mapping of 20cm GSD, meaning that each pixel represents a 20cm x 20cm tile in real life. Our 4cm GSD means that 1 pixel is 4cm x 4cm wide, allowing us to see and capture much more detail.
CGA: Sounds fascinating. So is this all there is to it?
Quaife: Almost. Because the photogrammetric software was developed to suit different industries data interoperability was an issue, we need to run the raw output of the capture process through a series of data format translations and post-processing stages to get individual models to be clean and solid. We invested a lot of time to develop these processes when creating the production flow line with the aim of making the models as end user friendly as possible across a wide selection of 3D modelling packages.
CGA: Could you tell us but more about these processes?
Quaife: I would prefer not to as these are part of our unique know-how that we try to protect.
CGA: No problem. As you described your flow line, all the processes seem to be very manual and time-consuming. Is there any room for automation at all?
Quaife: There are many automatically derived 3D models available, but we believe none of them are precise enough for our model. Automation tends to cut corners in terms of both accuracy and level of detail - to maintain our unique selling point we need to keep to manual capture by a handful of highly trained photogrammetrists.
CGA: What’s the most interesting or unique use of your models to date?
We have been lucky enough to work on a lot of exciting building projects within London, it’s fair to say we know how the London skyline is going to look in the next 5 – 10 years. However this year we have been involved in some great projects for the Olympics and Diamond Jubilee. We used our 3D model of Buckingham Palace to create artwork for the stage at the Windsor Royal horse pageant, which was televised a few weeks ago. It was great to see our hard work at the centre of such a prestigious event.
CGA:What’s the most surprising thing you’ve discovered through the process of analyzing and modeling London?
That even through difficult economic times London is constantly changing and evolving. It’s very difficult to model a building from the air when that building is under scaffolding!
CGA: Are there any limitations in capturing data the way you do? Are you able to capture everything 100% of the time?
See above! The main problem we face is the currency of aerial imagery available, we are limited to what imagery of London is available, if a building hasn’t been built when the image was captured, it’s impossible for us to model it using photogrammetry. This is when either a site visit is required, or we acquire the model from other sources or we have to exclude the building from the context model.
CGA: We recently posted a link to an article speculating about why Google sold SketchUp. How does what Google is doing compare to your own technology and processes? Do you envision they will ever start increasing their GSD or compete with what you are doing?
We always set out to create a highly detailed and accurate model as we knew that there are companies out there with far greater resources creating city models, however we believe that there is demand for both types of model. For planning purposes accuracy is key and the city models produced by Google and others aren’t of sufficient quality to fit this demand. The technology used by Google is certainly impressive and we are always keen to follow the developments made in this area, however I still feel that the process we use produces superior results.
CGA: For a while there photogrammetry seemed to take a back seat to new technologies like LIDAR but it now seems to be used a lot more often. Do you see this technique being superior or are there use cases for both?
There are definitely cases for both, for us Photogrammetry is the best tool to use right now for our needs. We do use LIDAR to scan building facades and merge this data with our own city model, we did this recently in and around Piccadilly circus and Regent street, the combination of both data sets produced a great looking model. I don’t think one method is proving superior right now, in my opinion whilst the capture of raw LIDAR data is quick the processing and storing of the data is difficult and costly and still requires operator input. Whilst photogrammetry is an old technology, with the advances in digital imagery and photogrammetric software it still produces the best results.
CGA: Your 3D model of London is a commercial product - who are your customers?
Quaife: GIA is still our largest customer but we work with many other building consultancies, planners and architects, however lately we have had many new customers from the press, TV, movie and advertising industries, especially with regards to the upcoming London Olympics and Diamond Jubilee. There was a lot of interest from the booming gaming industry as well, but so far we are not able to race around our model or blow parts of it to pieces.
CGA: So what is the future for Vertex Modelling? Where do you want to take it next?
Quaife: We still have some parts of London that need modelling and as you can imagine, this takes time to do. There is also systematic revision taking place where we continuously add, remove and change models as new developments are going up. There is lots of work going into upgrading parts of our standard LOD3 (Level of Detail 3) model to LOD4 - this involves adding facade detail to otherwise flat-sided photogrammetric models. We have pretty much finished the main London landmarks such as Tower Bridge, Houses of Parliament or Piccadilly Circus, but we will continue to upgrade our models to this higher level of detail for many more sites in London.
You must be logged in to post a comment. Login here.