Peter Murray spoke to our 2018 Technical Forum delegates about Brisbane City Council (BCC)’s use of Point Clouds in 12d Model software.
Peter works in the Surveying area of BCC, which is actually run as one large local authority covering the whole city; it’s unusual in that sense. It’s the largest local authority in Australia, by both budget and population (it covers a population of around 1.2 million and an area of 1,367 square kilometres). It works to an annual budget of around $3.1 billion – to cover traffic management and infrastructure, public transport, parks and opens spaces, economic development, and lifestyle and leisure. This leads to a great deal of variety in work – large projects and many smaller ones as well.
There are nearly 8,000 people working in the organisation. In Planning and Design, there are nearly 350 people – surveyors, road designers, drainage designers, bikeway designers, Geotech, pavement designers, landfill management, GIS, architects, landscape architects, water management, flood modellers, urban planners, and environmentalists.
What is a point cloud?
A point cloud is just a huge collection (thousands-billions) of points – they’re unrelated despite looking like they’re related. The enormous scale causes issues – bigger projects lead to more things that can go wrong. Point clouds have been in use since the mid-1990s…which gives some pause as to whether they’re still as relevant as they’re sometimes deemed to be.
How do we acquire point clouds?
A point cloud can be generated by laser scanning (e.g. Terrestrial, Mobile, Aerial), via photogrammetric techniques (e.g. UAVs), or using SONAR (e.g. Hydrographic Surveys).
Why do we use point clouds?
They appear to be very detailed and intuitive – they look almost like photographs (and it’s possible to measure between the points), which makes people think they’re loaded with useful information. Point clouds can also be captured rapidly and at a relatively low cost (particularly using LiDAR – this can lead to being able to capture, within days, huge amounts of information that would otherwise take surveyors years to collect). They allow for measuring of areas with difficult access, allowing for increased safety in areas such as freeways and dangerous industrial areas (e.g. through use of drones). And they’re intimately intertwined with BIM, which makes them unavoidable, especially as they become more and more mainstream and accessible to a wider audience.
What are the potential downfalls of using point clouds?
Highly specialised skills are required to produce a high-quality point cloud. The size of the datasets required is also prohibitive. It is also an issue that point clouds don’t fit well with traditional design processes, and are technology hungry – a variety of sophisticated equipment is required for their successful use.
How do we collect the datasets for point clouds?
- LiDAR – used since about 2010 by BCC for flood plain modelling, concept designs, investigations, and volumetric resumptions, LiDAR is regularly incorporated into their workflows and is an accepted data source. Data is generated by a plane flying over the ground with lasers pointing below and measuring downwards. As the lasers hit the ground, the beams are reflected back up to the plane so the plane can take measurements, at a rate of about 2 million measurements per second! Unfortunately, vegetation and water are the ‘natural enemies’ of LiDAR, so it can’t be used everywhere. Over time, their teams have learnt that not every point is required, nor is every point reliable…and unfortunately it isn’t always possible to pick the reliable points by inspection.
- LAS files (which have evolved from LiDAR) with categories, which are generated when reflections bounce off g. leaves and trees, leading to greater ability to filter out extraneous information, determine roof heights, interpolate floor levels, etc. LAS files are very useful in particular circumstances, and are included (with accompanying macros) in a number of BCC workflows.
- The new ‘Point Cloud Surface Thinning’ option in 12d Model 14 – a very neat function allowing draping of strings through point clouds. This means points can be concentrated where changes in grade occur, leading to a reduction in ALS points. The result is better-looking contours which are closer to the original…at 12% of the size of the original dataset.
- UAV LiDAR using drones – this can be great for working in what would otherwise be very dangerous areas.
- Mobile Laser Scanning – this is more accurate and less expensive than LiDAR, but also more difficult to control.
- UAV Photogrammetric Point Clouds – this method is quick and inexpensive, but again there are issues with control.
- Terrestrial Laser Scanning – BCC had experience with this years ago, for bridge scans. Using this method, small amounts of important information can safely be obtained.
What they have learnt at BCC
Overall, point clouds are an efficient and practical way of collecting a dataset. It is important to remember that not all the points are needed, and that not all clouds are the same. Also, file extensions are not a reliable indicator of contents – there are standards in existence, but they are not always followed. Peter also cautioned against such marketing claims as ‘Scan to BIM capability’ as they are not always what they seem.
Some of the point cloud outputs include a full point cloud (which is a good record of what was there), extracted objects, vectors and points, surfaces (TINs), and viewers. These can be used in such areas as geospatial, forensics, and film.
Point cloud functionality in 12d Model
Peter said there is definitely value in point clouds, but they’re not yet civil design ready, at least not universally. 12d Model manages point clouds well, though – it will read them in with ease.
12d Model will import common formats of point cloud, convert between formats, and perform projection transformations. It uses a ‘String_cloud’ element. In 12d Model 14, these processes have been improved even further – there is now capability to import multiple files, selected in Perspective view (which has also been made more responsive and reliable). Threaded views have also been added.
Peter also outlined some of his favourite point cloud functionality in 12d Model – including manipulating categories, deleting/undeleting, draping against point clouds, drawing flags, limiting clouds, pinning clouds, and of course the aforementioned Point Cloud Thinning.
Where to now for BCC and point clouds?
As they’ve now reached such a level of success with LiDAR point clouds, they’re now looking at scanning drainage chambers, scanning buildings, and data extraction and modelling (including vectors, trimeshes, and pipes). Peter showed examples of terrestrial laser scanning they’ve done (in particular with manholes). He has been investigating ways to utilise point clouds, including a macro (within 12d Model) to slice them, meaning he could extract a trimesh out of a point cloud to reduce it to a manageable number of points. By colouring the trimesh, surrounding spots have been made visible, and the clouds have become more valuable in his day-to-day work. By combining an image and a point cloud on some other projects, further usefulness has been discovered.
BCC has been developing a specification for the extraction of trimeshes from point clouds, as well as mapping files and 12d Field codes. They have utilised DTM auditing routines for trimeshes. Recently they purchased a BLK 360 scanner, and they are working on developing in-house skills to take their use of point clouds even further.
In essence, keeping full point clouds is a good way of maintaining an accurate record for future reference, and with some ingenuity, their day-to-day usefulness can be harnessed on some projects, too.
Watch Peter’s talk here: http://bit.ly/2PcZW2x