By James Creasy, VP of Engineering, SKUR
One of the unexpected pleasures of working with analytics on 3D imaging data is how extracting actionable results from an enormous pile of numbers seems like magic.
Unlike like a lot of data used in Big Data analytics, 3D point clouds contain almost no intrinsic information. Compared to, for example, airline flight arrival and departure data, the lines of data in a point cloud are as simple as three floating point numbers. No keywords that can be interpreted as a airport code, no numbers that represent time, no airlines names. Just several billion float values.
Here’s a short example:
Points in a point cloud have value because of their relationship to other points in the same cloud, or in relation to a design model. There are several types of relationships points can have with other points, here are a few:
- Nearest point
- Points within a radius of a given point
- Distance to a model surface
- Is the point part of a set of points that form a plane?
- Is the point part of a set of points that form a cylinder?
- Is the point part of a set of points that are a scanned object we want to ignore. An example would be a person walking through the scene.
Complicating this further is that a single point is usually not useful. If there is a fly crossing the scene during a scan, it may be represented as a single, and useless point. Generally we need to have several points in the relationships listed above to have any confidence there was actually anything there.
Once we understand the relationships, we can start to ask and answer interesting questions about complex scenes, such as:
- Is the top of a footing bolt in the correct position, despite being surrounded in rebar?
- Where is position of maximum deflection of a steel beam, 20 meters overhead?
- What is the exact distance of maximum deflection the beam?
- How far and in what direction was the access hole misplaced during construction?
- Is the As-Built model accurate? How do I correct it?
To me, it seems like magic to go from a list of lifeless numbers to accurate locations and displacements of real world objects. It’s a challenge I relish as part of the Engineering team at SKUR, where we process large point clouds and return actionable analytics for our construction industry clients.