Fast, accurate and automatic point cloud registration software
Terrestrial laser scanning has become a powerful way of capturing measured data about the built environment. However, the process to register the point clouds and extract useful information is often user-intensive and time-consuming.
Point clouds are created by laser scanners and can be used to create visual models of everything from geographic terrain to the built environment, or any object within them. Vercator software greatly improves the alignment of 3D point cloud data by making it a faster, easier and more robust process. Evolving from UCL spin-out technology, Vercator software lowers the barriers to onsite data capture, speeds up scan alignment and makes downstream analysis more convenient.
The software improves the time-consuming scanning phase and simplifies the reality capture process through:
These features deliver substantial user benefits through significantly increased cost and process efficiencies - less time staging a scan, less time extrapolating the data, yet still providing repeatable and accurate results.
Advantages of using Vercator software
Compatibility allows Vercator software to process most data created by static laser scanners. The software is compatible with major industry formats with more being added as it is developed.
Vercator software is compatible with most major industry formats with more being added as the software is developed
The Vercator approach
In the approach adopted by Vercator, vectors in overlapping scans are matched and brought into alignment, first to perform rotation alignment, and then translation alignment in the horizontal plane, followed by translation alignment in the vertical plane. In fact, such vectors are calculated and found at every point. Typically there are 10’s of millions of natural targets in each scan compared to the 10's of artificial targets or natural targets marked by eye in other approaches, resulting in fewer misalignments. The advantage of the Vercator approach is that the process utilises natural features in the 3D environment as natural targets which are automatically recognised, their location and orientation determined, then represented by feature vectors. A detailed description of the four methods can be seen below.
1. Details of the Method
Consider the natural features and targets to be small flat elemental areas, because many 3D environments being scanned have such features. Each flat element is represented by a vector direction, which is either normal or at right angles to each small flat plane, but its length is normalised to one. We can then draw these 'surface normal vectors' as small arrows starting at each point and pointing away from it, as shown in figure 1.
2. Rotational Alignment
Now imagine that each vector is lifted away from where it is, but maintains its direction and is moved so that its tail lies at the origin of a new space. All of the surface normal vectors have the same length so their arrow tips will lie on the surface of a sphere, as shown in figure 2. A pattern is created on the surface of the sphere. For example, a wall will have many surface normal vectors, all of which will be represented on the surface of the sphere, roughly at the same place since the surface normals will be generally parallel. This is carried out for each overlapping scan giving the sphere its own pattern for each scan. Now if the scans have sufficient overlap the resulting patterns will have many similar features. So, by moving the origin of the sphere for one scan to coincide with the origin of the sphere for an adjacent overlapping scan, we nest two spheres inside one another. Then we only have to rotate one sphere relative to the other until the two patterns match in order to obtain the angles we need to rotate one scan to bring it into alignment with the other scan. If the horizontal is known, it is only necessary to perform a rotation about a vertical axis until the two patterns match.
3. Horizontal Translation Alignment
To determine how much horizontal movement the scans require, the point cloud scans are projected, collapsed or squashed onto the horizontal plane to form a 2D plan view, figure 3. When the points collapse onto the plane, vertical walls which have millions of points on them will collapse to form a line on the plane. This creates a point density image. These 2D plan view images have already been rotated to have the same angular rotational alignment, so all that is necessary is to slide the image for one scan over that of the adjacent scan, then to calculate the degree of match to find the position of best alignment. The degree of match calculation takes into account the density of points.
4. Vertical Translation Alignment
To find the vertical shift or translation, the point clouds of the two scans are separately projected, collapsed or squashed onto a vertical rod, figure 4. Flat floors with millions of points on them collapse to very high densities of points on the rod, and similarly with flat ceilings. The point density pattern of one scan is slid over the point density pattern of the other scan to obtain the position of best match, which indicates how far one scan must be moved to bring it into alignment with the other scan. To ensure a flat horizontal floor in one scan aligns to a flat floor in the overlapping scan and not to a flat horizontal ceiling, the surface normal direction is retained in the process of collapse. Since all floor points have surface normal vectors pointing up and all ceiling points have surface normal vectors pointing down, the floor can be distinguished from the ceiling and only matched to points with surface normal vectors pointing in the same direction.
Vercator software has been developed initially for the construction industry where it is was used in a trial programme by BIM professionals and is now being rolled out to businesses. As the need to prevent and correct construction errors onsite becomes more pressing, the convenience of downstream data analysis with faster and verifiable information is sure to be welcomed by users.
With its ground-breaking technology advance, Vercator software users achieve state-of-the-art measurement accuracy and benefit through significantly improved cost and process efficiencies. This means less time staging a scan, less time extrapolating the data, yet still yielding reliable and accurate results
The software is also compatible with all common data sources and formats, making it simple to integrate into your current workflow.
Take a large building such as an office block recorded made up of over 130 laser scans. Registering the scans using conventional software took 28 hours under operator supervision. Vercator software automatically registered the same datasets in 16 hours, a 43% reduction in processing time without the need for operator supervision.
|Type of structure||No. of scans||Data size (.xyz)||Conventional alignment||Vercator alignment||Time saving|
|Large building (low resolution)||200||80GB||50 hours||18 hours||64%|
|Large building (high resolution)||130||244GB||28 hours||16 hours||43%|
|Street scene, central London||65||74GB||16 hours||4 hour||75%|
|Laboratory||28||34GB||5 hours||52 minutes||83%|
|Concrete building core, 1 floor||16||20GB||2 hours||14 minutes||88%|
Sign-up to your FREE trial of Vercator software and start registering point clouds automatically!
Read David Selviah's article
First published on 23.03.2018 in the RICS magazine and at the Geomatics World website.