Point clouds are datasets collected during laser scanning surveys. These data clouds consist of 3D (XYZ) coordinates that collectively represent scanned surfaces, detailing the lay of the land and the shape, dimensions, and size of topographic features and human-made structures.
Surveying teams use point cloud data to create digital models and maps of surveyed areas, allowing for data analysis to take place on workstations, rather than on-site. But before organizations can extract value from point cloud data, datasets need to be registered.
What Is Point Cloud Registration?
Point cloud surveys involve taking multiple scans of the same area from different angles. These scans allow organizations to turn point clouds into a digital model viewable from any angle. When hardware collects point cloud data, the XYZ coordinates that make it up act like pixels and form an image. Point cloud processing software takes these points to make a model, or map representing the surveyed environment. So with more coordinates from multiple datasets, a clearer image can be created and more information can be extracted.
Point cloud registration is a multi-step process aligning multiple, overlapping point clouds to form a detailed and accurate representation of the surveyed area.
The Point Cloud Registration Process
The point cloud registration process usually occurs using point cloud processing software or the manufacturer’s hardware-software solution. The right software will streamline the process, helping organizations achieve the best results as quickly as possible.
There are two main stages when it comes to point cloud registration: preprocessing and registration and alignment. Both stages are crucial to creating noise-free, clear, and complete point cloud models — and both are automated by point cloud softwares.
The first stage of point cloud registration involves removing noise and data outliers while preserving the clarity of features. This process can improve the clarity of the data and reduce the chance of outliers or noise interfering with data analysis.
During this stage, organizations can also choose to downsample the point cloud data sets (reduce the density of the point clouds by removing some data points) or delete data from unwanted features. This can speed up the point cloud registration process and create a more focused and concise dataset. However, for projects requiring a lot of detail from large areas, this isn’t always a good idea.
Registration and Alignment
Once point cloud data has been cleaned up and refined, these datasets are “stitched” together – with the registration software – using survey targets, control coordinates, common planes and/or notable features as points of reference. After this process, the point cloud data sets should be tightly calibrated, and the pixels should be aligned.
Factors Affecting the Point Cloud Registration Process
The point cloud registration process doesn’t always achieve the same quality of results. The quality of the stitched data and the speed of the registration process are dependent on factors such as data quality and the technology used for registration.
Targeted vs. Targetless Scans
Some organizations will choose to use targets during surveys. These targets are usually patterns of easy to identify reflective boards, used as points of reference when stitching datasets together.
The registration process with data from targeted surveys is often faster than with targetless data. This is because targets are quickly identified and matched. Data from targeted surveys can also result in more accurate image alignment. However, targetless surveys can be more common, as they’re more cost-efficient to carry out.
The larger the overlap ratio of different point clouds, the easier it can be to stitch datasets and achieve precise alignment. However, surveys will rarely gather data with a maximum overlap ratio as deadlines and budget restrictions make this unfeasible. An overlap ratio that is too low may be too unstable to merge point clouds, but often the registration process will need to be carried out with minimal overlap. High-quality results are still achievable, but the registration process may involve a few more manual steps — it can be more difficult to match the overlap, and additional features may need to act as control points.
Point Cloud Processing Software
The point cloud processing software your organization uses plays a big role in the efficiency of your point cloud registration process. Advanced software will automate the majority of the registration tasks and achieve more precise image and data alignment. Higher quality data always proves to be more beneficial in downstream operations.