What is a point cloud? Let UKAHT explain

02/04/2024

We use 3D point cloud models to bring our Antarctic bases from the bottom of the globe into the public sphere. But what exactly is a point cloud? UKAHT’s XR Producer Lesley Johnston explains.

Our historic bases are being digitally documented and virtually reconstructed to support our conservation work and make the sites accessible to anyone, anywhere in the world. We are using processes called photogrammetry and laser scanning to do this and a key stage in the reconstruction process is the creation of 3D point clouds. Some may even argue, it’s an art form in its own right.

Our XR Producer, Lesley Johnston, is using point clouds to bring UKAHT’s heritage into the public sphere and its stories to life. Point clouds are an important part of this process, but also provide exciting new perspectives on these old buildings in our care.

What is a point cloud?

The important thing to understand is that computers lie to us, and we let them. 

A computer doesn’t know what a building is, nor a ball, nor even a circle. However, computers do know numbers, and can position, or map, those numbers onto an axis as coordinates: left/right, up/down, forward/back, or, if you are a computer x, y, and z. Give a computer 2 coordinate points in space using the x and y axis and it can connect these and give you a line. Connect a third point and you have a triangle, a fourth a square and so on. 

Provide 20 points or more, and to the human eye, this collection of points joined only by straight lines, will start to look like a circle! The more points, the smoother the ‘circle’. Perhaps not quite a lie, but a striking optical illusion. Add in that third dimension (3D): axis z, and connect all the points again with a line and our triangle becomes a 3D mesh of a pyramid (tetrahedron), a square becomes a cube (hexahedron) and our circle is now a simple ball (dodecahedron).

Thinking in 3D (Credit: Shutterstock/Sasha701)

Simply put, a point cloud is a collection of points within three dimensions representing a 3D shape or object without any lines connecting them. When we ask the computer to show simple lines between these points we call this a 3D mesh, and when we ask the computer to colour the triangular or square panels between the points (we call these panels polygons), we call this a 3D model. In this case, the data points represent our historic bases and their surrounding terrain. 

Making a point cloud

Point clouds are often an unseen part of the virtual reconstruction process, with the final processing output most often being a coloured 3D model which can be used by construction or conservation experts as a planning tool for measuring or for display purposes (our team of heritage conservation experts are using this process to look after our Antarctic huts). Point clouds are created during the 3D reconstruction stage of photogrammetry and the capture or scanning stage of laser scanning which are commonly used methods to capture real-world objects and create these accurate digital models. 

A surveying station at Port Lockroy (UKAHT/Nathan Fenney)

The process of photogrammetry involves taking many overlapping photographs of an object, structure or space from many different angles around it and then converting them into a 3D digital model by feeding them into specialist photogrammetric software such as Reality Capture by CapturingReality. The software aligns those photographs in 3D space like a giant jigsaw, where it identifies the same features in the photograph's colour data, such as the corner of a desk or the edge of a wall, in two or more photographs it turns these into points with an x, y, and z coordinate which then makes up the shape of the 3D object being captured. This process can create highly accurate models, however, the accuracy depends on how many photographs were taken, the quality of the photograph, and the quality of the algorithm the software is using to make the 3D model, different software will give different results. A key benefit of photogrammetry is the colour data which is provided by each of the photographs.

The result: triangles are where photos were taken; cubes are where the surveying stations were when each laser scan was taken (Credit: UKAHT/Lesley Johnston)

Laser scan point clouds are created when the laser scan is undertaken on location. Laser scanners project a single laser beam out in all directions from one rotating central point, recording the angle on the three axis (x,y, and z) and the time the laser beam takes from being sent to hitting an object to map the surrounding environment. This gives a very mathematically precise point cloud although this process does not provide good colour information for the point cloud or 3D model as laser scanners only take one large 360 photograph. Take a look at our UKAHT point clouds and see if you can tell which ones were laser scanned and which ones were made only by photogrammetry, it won’t be easy.

Using either method, the more points that can be mapped by providing the photogrammetry software with more images, or undertaking laser scanning from more positions, the more detailed and accurate the point cloud will be. Our conservation teams returned from Antarctica with dozens of laser scans and thousands of photographs from digitally capturing only a few of our bases in previous seasons, this came to around 200 gigabytes of data before we began processing the data.

The most thorough point clouds are created by combining both of these two processes, improving the accuracy of the laser scan and the colour information from photogrammetry photographs. It needs to be noted that the more information input to the creation of a point cloud, the more data added, the bigger the file output size will become. This presents challenges as larger files require specialist software and more advanced computers to open or view the file. A simple or sparse point cloud can be a way to view the object that the point cloud represents.

Lesley at work at Detaille (Credit: UKAHT/Lesley Johnston)

By using specific workflows we can export the point cloud from the reconstruction process and share it publicly with 3D viewing platforms such as Sketchfab. 3D viewing platforms such as Sketchfab and Potree have begun making it possible over the last decade for point clouds to be shown both for their unique visual appeal and to show the accurate topography or surface shape of items or landscapes that would be too large a file to share publicly otherwise, our own point clouds have been simplified considerably to make them a suitable size to share online. 

Point clouds can be a project or output themselves as they provide new perspectives on space whilst sharing the important features of a building or object, or in the case of our amazing Port Lockroy site, the sheer size of the island and all the manmade features that helped scientists survive and work in such a beautiful but inhospitable landscape. Point clouds can also be considered an art form by themselves, an idea that has also been explored in the book The Art of the Point Cloud.

Point clouds to polygons

Point clouds are an essential part of the digital reconstruction process. However, they are not typically the end product. 

Point clouds, as multiple coordinates in space, are the core structure that 3D models are built on. In the following processing stages, as noted above, the software then calculates the connections between the points creating a 3D mesh creating flat panels between the points. Think of this as a simple football, all of the leather patches of the ball create the 3D mesh of the object, and each patch is called a polygon. These connected polygons make up our 3D model, and just like our circle example above, if we add more points in the point cloud then there are more polygons and the more spherical and realistic the ball looks. 

The final piece of work is to add colour to make the object look realistic. This is texturing, the software takes the colour from the many photographs that made up the 3D jigsaw of the point cloud and applies this to the polygons to make the object look fully realistic. 

Point clouds are used for a range of purposes including creating 3D computer-aided design (CAD) models for manufactured parts, for metrology and quality inspection and for various visualisation, animation and rendering applications. 

Point clouds are particularly useful for mapping heritage sites in hard-to-reach or dangerous places. A good example of this is the Ottoman palace of Beit Ghazaleh in the UNESCO-listed Ancient City of Aleppo, Syria. The site suffered substantial damage from warfare and earthquakes in the region. A high-precision 3D survey of Beit Ghazaleh was conducted in 2017 which resulted in a 1.2 billion data point cloud render of the heritage site.

The Beit Ghazala Point Cloud Survey (Credit: UNESCO)

Renders such as this can be used for documentation, restoration, conservation and research purposes. In terms of restoration and conservation, 3D models can be used to detect structural weaknesses, corrosion and defections without the need for specialist teams to physically be at the site. As such, complex restoration projects can be planned from the safety of a computer screen.

From point cloud to virtual reality

In 2016, a conservation programme was initiated to create detailed site surveys and 3D base models of UKAHT sites with the Mapping and Geographic Information Centre at the British Antarctic Survey (BAS).

During the 2019/20 season, our team began capturing data from our Antarctic bases, laser scans and photographs to reconstruct the bases digitally. UKAHT have been working with Anglia Ruskin University (ARU) and ARU Storylab since 2022 to turn these 3D models into a new and exciting project called Immersive Antarctica which includes a virtual reality (VR) experience.

The Immersive Antarctica showreel (Credit: UKAHT/ARU Storylab)

Through the Immersive Antarctica programme, we have already virtually rebuilt key areas of Base E, Stonington Island, to an approximation of its condition in 1965 using photographs for reference and a little bit of artistic licence to recreate the base as it was when it was an active site. 

Using the same process, we will do the same for our other sites on the Antarctic Peninsula so people will be able to experience our bases – 9,000 miles away and 80 years ago – from the comfort of their own homes.

Point clouds of our sites

Use the embedded 3D point cloud models below to explore our sites or visit our Sketchfab homepage.

Base A, Port Lockroy