Public engagement

Bentley's OpenCities Planner and ContextCapture software applications provide an ideal environment for developing projects that seek greater engagement with the public, as Gerpho 3D's Trappes project in France shows

Bentley System's ContextCapture software and its OpenCities Planner have been used to develop a public project in Trappes, France, which encouraged bidirectional engagement with its citizens. The project was recently featured in one of Bentley's online webcasts, and gave a fascinating insight into both the background of the ContextCapture application and its subsequent use with OpenCities Planner, used in this instance to present a potentially controversial project to a city's inhabitants, and to invite their feedback.

The city was Trappes, South West of Paris, France, and presentation was given by Philippe Graindorge, founder and CEO of Gerpho 3D, an aerial photographic company. Philippe played a key role in the development of Bentley's immersive reality tool, and gave an illuminating demonstration of its origins and early development, with some practical guidance on how it works, and how to get the most out of it.

As an aerial photographer, Philippe was commissioned to produce high quality images of France's wealth of historic buildings and picturesque countryside. through the course of his work he came across an application called Acute3D, developed by a couple of mathematicians who, as it happened, were alumni of Philippe's own university.

Acute3D is able to convert a series of photographs of a group of buildings, or a stretch of countryside, into accurate 3D models, and with a common interest Philippe contacted and subsequently went into partnership with its developers to further develop the application. Acute3D was eventually acquired by Bentley, becoming ContextCapture in the process.

How does it work? The concept is pretty straightforward. A series of photographs are produced using standard professional quality cameras, gathered together in a file and uploaded to the application. The software deduces how the 3D object you are looking at is made and creates a 3D mesh of it - a field of triangles that define the building and its surrounding environment. The software then goes back to the photographs and applies the colours of the photos on to the mesh and voila, there you have the 3D model.

It's an impressive piece of software that makes the process look easy, but it requires a powerful processor and the latest graphics cards to process the math routines behind the 3D reality mesh.

That all depends on the number of photographs taken, the size of the model, which can be hundreds of square kilometres, and the accuracy required. This is determined by the resolution of the camera use and can be in millimetres or as small as 1cm per camera image pixel. Model sizes can therefore be huge and in the hundreds of gigabytes, and so to view even large models in real-time a LOD (level of detail) structure was built into the software, which divides the model into much smaller files, at different resolutions, dependent on its purpose.

The calculations behind creating the models are intense, and the processing power and graphics capabilities required for the largest models need to be as good as you can get. Even then processing the data could take hours, days or even weeks. Originally, ContextCapture users were provided with a dedicated viewer, but the application has since evolved to widen its use by launching a 3D viewer, which enables any user, on any device, to access and navigate through a ContextCapture model.

So far, so good - but early on in the development of the application, Philippe at Gerpho3D was asked by a local council to provide a composite model of a historic chateau, created using ContextCapture, and to add a CAD model of a proposed road passing close to it, to ascertain viewing angles and confirm that the scheme would not impinge upon the chateau's visual impact.

His demonstration was successful, and the road was passed, but the process of adding the CAD model in its early stages involved some experimentation with the different formats, which meant that any subsequent and similar demonstrations took time to set up, and required 'DIY' solutions to accomplish.

The problem was eventually solved in 2019, when Bentley updated a previously acquired application and renamed it OpenCities Planner to simplify the process of adding CAD models directly into ContextCapture immersive reality models, making complete models available online for everyone, in high resolution irrespective of the expertise of the viewer, or the viewing device used.

As Philippe explained, creating immersive reality models was a top-down exercise, originating simply as a way to inform the public. What excited Philippe was that it provided a means of communicating directly with the public and inviting their feedback on proposed developments - and allowing it to become a bottom-up process. Previous planning proposals were presented using 2D drawings and static architectural sketches, which meant little to the majority of the public with no technical expertise.

To develop the concept further, the town of Trappes asked Gerpho3D to create a model of the major road scheme that cut through the town, and to provide a means of encouraging feedback from its citizens. Accordingly, ContextCapture was used to create the virtual reality model of the town, in enough detail for local residents to be able to identify their homes, the local schools, shops and other features.

Using the architect's blueprints, a detailed 3D model of the road scheme was created and integrated and the whole model placed on the town's website.

Visitors to the website were able to click on the model and view the project in different modes, before or after construction, from their own viewpoint. Different layers of data were established to furnish the model with people and cars, trees and street furniture and other features to enhance its visual appearance. They also provided the ability to view additional details for those who wanted more technical information about some aspects of the project, points of interest within the town, and any other relevant data.

To satisfy the two-way requirements of the model, Form Builder, a standard function in OpenCities Planner for interacting with citizens and stakeholders, allows the project to define survey questions and relate those to the interactive 3D scene. This dialogue function is a central function for engagement and involvement in urban development, or to invite interaction with external users around safety, fault reporting, statutory or non-statutory consultation, or to open up widely for opinions or feeling in connection of events or broad city aspects.

Another important feature is the rights management tool, which allows different levels of access to the model, from developers to site visitors, enabling all users to engage with the virtual model at their own level.

What could be simpler? You don't need to import any software, as it is all handled within Microsoft's Azure Cloud. Bing's mapping tools provide the basic maps and satellite data, from which the designated area is cut out, and the reality model created by ContextCapture is imported and merged with the map. CAD models can then be dragged and dropped into the environment, either georeferenced for precise positioning, or non-georeferenced for other CAD elements like cars, which can subsequently be moved into place.

Philippe also shared some useful tips within his presentation to guide others using ContextCapture in response to questions from the audience. When preparing to take photographs, use a handheld rather than an automatic camera, as it allows the user to focus on the best shots that can be taken to create the perfect 3D virtual reality model.

Additionally, users should be prepared to spend some time after the model has been processed in touching up parts of it, using Photoshop or other tools. Shadows are also important for the veracity of the model, and tools allow dates, times of year and weather data to display models in any prevailing condition - with the shadows generated accordingly.

Finally, reflective areas like lakes and ponds don't work well with photogrammetry, so workarounds have to be made to ascertain water levels and render accordingly.