This is part 3 of a multipart blog series on the Virtual CTO project: A digital reconstruction of the CTO measurement train commissioned by the section Railway Engineering at Delft University of Technology. In part 1 of this blog series I covered the process and results of making LiDAR scans of the train, both outside and inside. In part 2, I described the CAD modeling process, and in this part I continue with describing the process and showing results of making handheld 3D scans of various parts of the train chassis, bogies and pantograph.
Handheld 3D scan of skirts and beams fitted to one of the Y32 bogies (isometric view after point cloud cleanup and merging parts in VXscan)
The idea to make additional handheld 3D scans on sections of the CTO measurement train was born after it appeared that some (custom) parts were not described in construction drawings and were also difficult to capture with a lidar scanner.
PREPARATION
I started preparing for handheld 3D scanning early 2018 by doing research and experiments. In the following paragraphs I briefly discuss my scanning workflow and some of the challenges I faced:
Decide which scanner to use
Initially I considered various handheld 3D scanners, but as the TU Delft appeared to have a Creaform HandySCAN 700 available in their CEG lab, the choice was quite easy. With its superior scanning resolution, accuracy and workflow, there was no need to look for an alternative solution. The following workflow features stood out to me:
user-friendly scanning environment VXscan that simplifies the scanning process,
fast workflows for scanner calibration and configuration (needed before every scan),
reflective positioning targets applied before 3D scanning, either on the object or its immediate surroundings, ensure a solid reference point for the 3D scanner,
realtime visualization of the 3D surface on a laptop as the object is being scanned,
scan in cross grid mode for faster data capture, while the additional single beam mode is used to go deeper into pockets and harder to reach areas,
shutter speed can be adjusted without stopping the acquisition process, which is sometimes needed for better performance on dark or reflective surface sections,
an optimized mesh is automatically created and available upon completion of the data acquisition step.
Creaform HandySCAN 700 handheld 3D scanner
| Technology | Light source | Recommended part size range |
Scanning area | Stand-off distance |
DOF | Resolution | Accuracy | Dimensions | Weight |
|---|---|---|---|---|---|---|---|---|---|
| Laser triangulation | 7 laser crosses (+1 extra line) | 0,1 - 4 m | 275 | 300 mm | 250 mm | 0,05 mm | < 0,03 mm | 77x122 | 0,85 kg |
Creaform HandySCAN 700 specifications
Choose the scan resolution
The most important parameter that needs to be set, after scanner calibration and configuring the scanner for the type of scanned surface are done, is the scan resolution. The scan resolution can be set independently of the size of the scanned object, but it greatly affects the amount of detail in the scan output, as well as the file size and the time it takes to create the optimized mesh. The resolution can be changed at any time before or after the scan.
In order to get a better feeling for the effect scan resolution has on the detail visible in the scan output, I made a series of test scans of a View Master for a range of available scan resolutions.
In the figure on the right you can see that the finest detail is disappearing between a scan resolution of 0,2 and 0,4 mm. But more importantly, for objects of this size, the scanned shape already starts to deviate significantly from the original, for scan resolutions above 1 mm.
Test scans of a View Master showing the effect of scan resolution on the accuracy of the scan
The View Master test is also a nice example to demonstrate, that for some objects (in this case placed on a stand), it’s also allowed to move and rotate the object during data acquisition, meaning you can scan all sides of an object in one go. As long as at least three targets remain visible from the part of the object that is already scanned, new targets are detected and registered on the global positioning model as the scanner is moved around the part.
FIRST SESSION
The first scan session took place in February 2018 in the CEG lab at Delft University of Technology. The object to scan was a duplicate of the Faiveley AM56 pantograph, that is also mounted on the roof of the CTO measurement vehicle.
Particularly challenging for this scan was the large amount of thin pipes in the structure, that became even more challenging when the pantograph was in its upward position. What can happen in a case like this, is that somewhere along the way the scanner looses track of its current global positioning model, and creates a new misaligned starting point that causes a ‘breaking point’ in the final scan. This is exactly what happened at one point with the scan for the upward position.
Because of the limited time available for scanning, the resolution was set to 1 mm. For an object of this size, automatically creating an optimized mesh upon completion of the data acquisition step, can take quite a while using high resolutions (depending on the hardware VXscan is running on).
Scan setup with Creaform HandySCAN 700 controlled by VXscan on Windows laptop
The final scan for the downward position came out fairly well (see below), but in hindsight, it would have been better to place the pantograph flat on the ground without the wooden support beams. That way additional positioning targets could have been applied on the floor, providing a stronger base for registering new scan positions to the global positioning model. Also, choosing a slightly better scan resolution would probably have made the scan process a bit less sensitive to leaving holes in the scan.
Handheld 3D scan of the Faiveley AM56 pantograph (after point cloud cleanup in VXscan, but before mesh cleanup in ZBrush)
SECOND SESSION
The second scan session took place in April 2018 inside in a maintenance workshop at NS Train Modernisation in Haarlem, in the same period in which I did the second lidar scan session. Because the bogies were in maintenance, I had perfect access to the entire chassis and lower side of the body, as well as the bogies themselves. The parts I focused on included (custom) parts of the bogie, that were not described well in construction drawings, as well as some parts of the chassis that showed blind spots in the lidar scans.
One of the most challenging scans I did, was for the custom skirts and beams fitted to one of the Y32 bogies, as shown in the image at the top of this blog post. This scan was done in two sub-scans (see image below): I made one scan for each side of the bogie with enough overlap in the middle to be automatically merged within VXscan afterwards. To be able to scan both sides of a skirt in one go, positional targets on the ground close to the skirts played an important role. The scan resolution was set to 0,8 mm.
Handheld 3D sub-scans of skirts and beams fitted to one of the Y32 bogies (isometric view with positional targets before point cloud cleanup and merging in VXscan)
Another interesting one I did for the chassis, was the scan for the buffers and the tow bar, as shown in the images below. The scan was done in three sub-scans, one for each of the left and right buffer areas, and one for the middle section with the tow bar. Again, the overlap between scans made it possible to merge them automatically afterwards. The scan resolution was set to 0,5 mm.
The optimized meshes in the second session turned out to be of really good quality, despite the challenging shapes. Besides using a slightly better scan resolution, the most important difference with the pantograph scan approach was my increased awareness, that positional targets are needed mainly for an accurate registration of the continuously changing scanner position, and not necessarily for every surface that needs to be captured accurately. The outcome was a scan behavior where I not only monitored the stand-off distance and a preferably perpendicular scanner orientation, but especially also which positional targets were in sight and whether they were already part of the global positioning model or not.
Pattern of ⌀6 mm black contour adhesive positioning targets
Handheld 3D scan of the CTO buffers and tow bar (after point cloud cleanup and merging various parts in VXscan, but before mesh cleanup in ZBrush)
After the scan clean up and merging was done, I aligned all handheld 3D scans with the lidar scans, to have everything in the correct position in the same 3D space, ready for further processing and remodeling. More about this in a future part of this blog series.