By James Creasy, VP of Engineering, SKUR
Most existing 3D modeling and image tools are designed to assist a human in making analytic judgments about source data. There are tools to smooth the noisy output of LiDAR scans, convert points to 3D meshes, and provide lighting effects to help a human understand a point cloud better.
The future is that computer algorithms can and will do a better job than humans in making analytic judgments from 3D image data.
In recent years, new types of computer algorithms are able to perform real-time judgment-based activities long thought to be the sole domain of humans. Examples are self driving cars, AlphaGo’s astonishing win over a professional Go player, and especially the plain-looking Google search results. The common theme here is that the new class of machine learning algorithms is poised to disrupt judgment-based tasks.
What does this mean for professionals and companies in the 3D imaging space? Primarily that 3D imaging data will be thought of more like any data source, another form of Big Data, rather than as an intermediate form to lighting pixels on a computer display for a human to squint at.
If you’ve spent much time with point clouds, you know they can be difficult to interpret, cumbersome to transport and view, and filled with visual artifacts that obscure what’s important. As mentioned above, there are many tools to help improve the visual appearance, and even tools to help calculate volumes and distances, but by-and-large they are operated by a highly trained person. This person knows how to combine the various tools to get the best possible visual display to make better assessments.
But this is not the future. Here in the data science team at SKUR, we view 3D point clouds as a data source for Big Data analytics and apply modern data science principles to them. The result is our analytics not only automatically do much of the work of cleaning and interpreting the 3D data, we also use modern techniques to extract actionable analytics which are then presented in a simple and straightforward way, no more squinting!
Relieved of the constraint to create a result limited to the capabilities of the human eye, our analytics reveal truths in the data invisible to human sight. A recent example was determining the positioning of anchor bolts in a complex scene where the bolts were already 80% sunk in concrete and surrounded by steel reinforcing rebar. As hard as we tried, we could not even see the bolts in a viewer. However, the SKUR computational engine was able to determine the placement of the bolts to within a few millimeters.
The fast approaching reality is that modern Data Science techniques are going to revolutionize many industries, in our case, producing high-level, judgment based analytics from 3D imaging data.