Using Software to Identify Geometric Shapes in Real Time
Erin Green posted on April 19, 2016 |

One of the most time-consuming steps in 3D scanning and data manipulation is translating information from a point cloud into usable 3D data like a mesh. But what if a software could skip this step and go directly to an object’s 3D geometry—in real time?

One of the exhibitors from the SPAR 3D Expo and Conference 2016, CurvSurf, has a software offering that can reportedly do just that: recognize and measure 3D object geometry in real time using a run-of-the-mill scanner (think the Xbox Kinect) or a Project Tango-enabled tablet.


Finding Surfaces with an Xbox Kinect

In the gaming industry, sensors such as the Kinect or even the Nintendo Wii’s controller can be fairly limited in their capabilities. It’s not uncommon for them to misinterpret a gesture altogether, often resulting in gaming missteps.

This penchant makes it quite fascinating that at SPAR 3D, CurvSurf’s FindSurface and FindCurve software set was paired with the Xbox sensor as a demonstration of monitoring a perpetually shifting geometric shape.

The team’s demonstration involved a cardboard box mounted on a rotating stand which was being monitored and measured by the software. FindSurface is designed with a geometric abstract-based algorithm which estimates a shape’s proportions based on a defined and recognizable shape—such as that of a box.

The software works with the sensor to recognize a shape based on a gazing point. Here’s a look at how it works:


Finding Surfaces with Project Tango

FindSurface in use on a Project Tango tablet to identify various shapes. (Image courtesy of CurvSurf.)

The second platform used to demonstrate the software’s capabilities was a Project Tango development kit. Project Tango is an endeavor from Google to bring computer vision to mobile devices, which perfectly suits the CurvSurf mission to bring point cloud identification to phones and tablets.

Using the Project Tango tablet, the team demonstrated the software’s ability to identify various shapes based on its gazing point and parameter estimation.

After identification, shapes can be exported with their XYZ data, radius and centering information into formats like STL and DWG.


AEC and Machine Learning

Although it isn’t clear whether FindSurface and FindCurve can extrapolate exact dimensions based on shape recognition, the software has some applications for both building monitoring and machine learning.

The software is in fact designed to parse its shapes from point cloud data almost instantaneously, which holds the potential to simplify large-scale scanning projects and create usable models for engineers and architects alike in much less time.

It also holds potential for machine learning, where it could eventually provide robots and computers with the ability to recognize certain shapes relevant to their tasks. However, as the software can currently recognize only six individual shapes, this is still a slightly futuristic ambition.

FindSurface and FindCurve are both commercial products, sold together for approximately USD$3,000.

For more information, check out the CurvSurf website.

Recommended For You