The Future of 3D Capture Technology?
- anthony milou
- Jul 16, 2022
- 9 min read
Updated: Nov 3, 2024
An overview of what technologies are currently at the bleeding edge of what is possible and worth keeping an eye on in the hope that they will lead to improved 3D capture results
Beyond Terrestrial 3D Laser Scanning
Terrestrial 3D laser scanning has long been the gold standard for 3D capture with a level of accuracy and detail capability that is still considered unrivalled in relation to any other capture method currently available. The advantages however are basically due to the fact that terrestrial 3D laser scanning relies on completing individual scans from fixed positions and then stitching all the scans that are completed into a single 3D capture. After each scan you need to physically pick up the scanner and the tripod it stands upon and move to the next location and so on until you cover the entire project site.
There is only so much you can cut out of the actual scanning time with incremental future improvements, while the time it takes to move the scanner from one scan location to the next is more or less a fixed time factor that cannot be drastically reduced. You could conceivably mount a terrestrial 3D laser scanner on a robot and have them autonomously do the work for you by moving from one fixed scan position to another. In fact this is what Leica has done by partnering with Boston Dynamics as you can see below and there are obvious benefits here in terms of automation, however on the topic of dramatically reducing the total amount of time in the field such a solution would not greatly alter the amount of time it takes to complete the task out on site.
It can therefore be concluded that the limitations in terms of how much area within a set amount of time could be covered by a terrestrial 3D laser scanner opens up the door for other technologies to eventually take over.
3D Laser Mapping - Hand Held or Worn Systems
It has often been cited that the time it takes for a 3D capture with a mobile mapping system can be as much as 10X less when compared to the above traditional method of terrestrial 3D laser scanning and in my direct experience this is basically correct (actually some projects I have completed with mobile mapping that were also previously completed with traditional scanning have shown a far greater than 10X reduction in time spent on site).
You can see a video here that gives a good idea of just why the reduction in time is so great based on the equipment and resulting capture process differences between the two technologies as you move through a project site:
By no longer relying on fixed scan positions as with terrestrial 3D laser scanning and instead focusing on continuous 3D laser mapping as you move through a project site (i.e. SLAM - Simultaneous localisation and mapping) the time saving benefits are obvious.
One of the main lagging factors up until now has been the accuracy of the captured data, though as I have written in another post there has been a lot of progress, with accuracy that can now rival that of terrestrial 3D laser scanning. With this post I focus on giving examples using the NavVis VLX V2 as I currently own this system and it is a good example of the current mobile mapping state of the art when it comes to the accuracy and level of detail possible for 3D capture.
“With this particular project the mobile mapping was completed in just under one hour while the terrestrial scanning was completed over two days”
It is important to mention that the level of detail that can be captured is quite high with detail at the level of 6mm of resolution possible with the resulting 3D pointcloud model. I would expect hopefully within the next 5 - 10 years to see improvements with 3mm of detail possible with the end results.
You can see below a side by side comparison between the Leica RTC360 terrestrial 3D laser scanner on the left and the 3D capture results of the NavVis VLX V2 mobile mapping system on the right (if interested you can also take a look at many more side by side comparisons for this project at the Dropbox link here).
With this particular project the mobile mapping was completed in just under one hour while the terrestrial scanning was completed over two days.
Other than dramatically reduced capture times, I have also noted improved visual quality of the colour application in the 3D capture. With the below example this is even more clear; even though both examples are pointclouds of the same area, due to the continuous mapping nature of the NavVis VLX, the colour application is smoother and the overall capture appears cleaner and easier to discern all the details present as seen on the right. The improvement of how colour is applied actually helps make up for to some extent the reduced resolution of the NavVis VLX (6mm) compared to the RTC360 (2mm).
One last observation with the above project is that there are few if any gaps in the 3D capture from the NavVis VLX compared to the RTC360. The reason probably comes down to the fact that the continuous mapping from the VLX covers more ground and less chance of missing a spot while on site. With the RTC360 there will inevitably be some missing data as there will not always be complete overlap and coverage from all the fixed scan positions taken (especially with very cluttered environments), while you do in fact have the luxury of moving through and covering almost every square metre and behind objects when mapping.
3D Laser Mapping - Drone LIDAR Systems
Drone LIDAR based mapping is improving all the time in terms of the accuracy, resolution and cost of the systems to purchase. Drone LIDAR systems are currently typically designed to work over large outdoor environments. Something to watch out for is the possible availability of a small form factor drone with the level of accuracy and resolution similar to the NavVis VLX mobile mapping system. Such a system would be very useful for out of reach or hazardous to capture areas ( i.e. an inaccessible roof of a building).
Below is a direct comparison of the current capabilities for laser based 3D capture technology.
From left to right you can see the level of accuracy and a visual of the resolution possible for terrestrial laser scanning, mobile mapping and LiDAR drone mapping :
If interested in further information regarding LiDAR drones a good place to start is with the videos by "Indiana Drones" where there are a lot of helpful comparisons made between the different Drone Lidar options:
3D Laser Mapping - Autonomous Systems
Even with mobile mapping as described above, having a person in the loop is probably the final limitation in achieving faster capture times and potentially better quality coverage of the project site.
Ultimately at some point in the future you should be able to turn up to a project site, from an app on your phone geo-fence the area of the property you would like captured and simply have the task completed with the building and property inside and out documented at a level of accuracy and resolution equal to current terrestrial 3D laser scanning.
We are not quite there yet when it comes to all the critical factors of hardware size, accuracy, resolution and cost. Reaching such a milestone of having just the right product could take 10 years, but a lot of progress in this direction has already been made. An example is the Emesent Hovermap and also the recent release of the Leica BLK2FLY and BLK ARC, all of which are capable of autonomous 3D capture.
Other than LiDAR, all of the above 3D capture solutions feature on board cameras and there is a lot of innovation happening at the moment in terms of using photographs and video to generate 3D models. It is possible that the photos taken could help augment the final 3D LiDAR data beyond simply assisting to colour the final scans. One good example is the recent progress with Instant Neural Radiance Fields from Nvidia (or NeRF for short). Instant NeRF allows just a few photographs or a short video to be used to generate a 3D model complete with reflections and the capture of clear surfaces and it computes this in a matter of a couple seconds! Instant NeRF gives some idea of how AI and improvements with computer processing power in general is also opening up new possibilities for 3D laser scanner hardware to potentially do more with the processing of the raw laser and image data captured.
An example of an Instant NeRF, in this case of the university of Otago (note the reflections in the water, credit to Stefanie Zollman):
Something that until recently was impossible with 3D capture; the capture of glass and reflective surfaces! Note the transparency of the surfaces in this Instant NeRF 3D capture of a Ferrari and of the windows of the building in the background. (credit to Johnathan Stephens):
Another instant NeRF 3D capture; Eze France (credit to Johnathan Stephens):
A very good overview & explanation of how the Instant NeRF AI works here:
The catch based on my own experiences with the Instant NeRF AI software to date is that conversion to a 3D mesh model doesn't yet produce an acceptable overall mesh result as you would have with photogrammetry, yet its highly possible such AI tools or similar could one day help with 3D laser scanning so that the AI can fill in the gaps and model via post-processing surfaces such as glass or mirrors which are currently not possible due to the inherent limitation of LiDAR.
4D Capture - Dynamic & Live Interaction Possibilities
Ultimately thanks to improvements in AI, it's only a matter of time before computer processing of 3D laser scanning data reaches the point where all surfaces captured of a building and its surrounds are recognised for what they are (i.e. windows, doors, floors, walls and individual fixtures and fittings etc) to automatically create a 3D Building Information Model (BIM) of individual parts that would be indistinguishable from one that was crafted by hand from scratch.
Eventually, it would not be surprising if the time between capture and BIM generation was so minimal that you could have real-time interaction between the physical and virtual 3D model created; or "live" 4D BIM environments to work with.
Imagine the possibilities for urban planning, design and construction if instead of a static 3D model you could observe and design within a dynamic environment that reflected what was actually present and happening right now in the real world. Currently the benefits of 4D BIM are limited due to the lag in time between capturing data on site and processing this into a completed model. Reducing this lag (i.e. with faster mapping such as with the Navvis VLX) and eventually achieving real time or live capture and modelling, will open up the true potential of 4D modelling to reflect a more direct representation of a project site.
To reach this point you would need real-time 3D capture which for the time being at least is currently only being applied in the market within very limited use cases such as with the Leica BLK247 intended for surveillance purposes or as is the case with LiDAR for autonomous navigation such as the Velodyne Velarray M1600 near-field solid state LiDAR.
The current limitations include the current low resolution and accuracy of the hardware compared to terrestrial 3D laser scanning, limited network bandwidth to transmit what would be enormous amounts of data and the processing power required to convert all this data to a real-time 4D BIM environment to work with. It took 20 years to go from the first commercial 3D laser scanner with the Cyrax 2400 to the current state of the art in 3D laser scanning; expect a greater level of progress in the years to come.
Summary of Expected Milestones
The expected 3D capture breakthroughs as described above can be summarised as follows:
1: 3D Laser Mapping on Par with Terrestrial Laser Scanning
3D laser mapping with better than 6mm of accuracy (i.e. the current capabilities of the NavVis VLX) in order to finally be on par with the capabilities of terrestrial 3D laser scanning.
The optimist's timeframe:
Within the next 3 years
2: Autonomous Drone LiDAR Mapping on Par with Ground Based Mapping
Drone LiDAR mapping with better than 20mm of accuracy (i.e. better than the current capabilities of the Emesent Hovermap) in order to be on par with the 6mm accuracy capabilities of the NavVis VLX V2 mapping system. In other words something at least as capable as the NavVis VLX so it can take to the air, combined with a small form factor and an accessible price point.
The optimist's timeframe:
Within the next 5 years
3: 4D Laser Capture & Real-Time BIM for Urban Planning, Design & Construction
The availability of real-time 3D laser scanning on par with current terrestrial laser scanning capabilities in terms of accuracy and resolution, along with the required AI computer processing capabilities in order to allow for 4D Building Information Modelling (BIM) to become ubiquitous.
The optimist's timeframe:
Within the next 8 years
Comments