Robust LineFit3D with STC-0038

When working in 3D – basic operations like line, plane and cylinder fitters are extremely important functions. Based on Scorpion Vision unique component technology we have upgraded the STC-0038-FitLine3D with the option to do a very robust 3D line fit.

Configuration Screen for STC-0038-FitLine3D

Works both in 2D Images or in 3D Points Cloud

The sample is a 3D stereo vision system where we show how you can pick pipes from a storage.

The pipes are located in 3D with Scorpion 3D Stinger technology

Standard LineFit3D with OpenCV algorithm

We do a linefit using the eight 3D points that are the center of the pipes using the OpenCV algorithm.

The OpenCV algorithms uses all points to create the 3D Line

Robust 3D Line Fit – Theil Sen Algorithm

The robust line fit evaluate the incoming 3D points and based on a median fit assumptions it removes the outliers – in this sample it concludes the three upper points is not part of the line – which is a good conclusion.

The robust line fit treats the three upper pipes as outliers – the longest / best lines in the points of the lower pipe row. The yellow points shows the incoming points

Conclusion

The Theil Sen algorithm for line fit is a good addition to the reliable 3D tools of Scorpion Vision Software. The robust median will help remove up 29.3% of the outliers and still make a good estimate. This feature will help make applications reliable. It also helps us to be significant better than OpenCV which is a requirement for us 🙂

The 3D fitted line in the Scorpion 3D View

There is also a 2D version of this algorithm implemented as a Scorpion Tool Component – STC-0040 FitLine2D. This component can be used to locate a line in a noisy environment.

A large range of Scorpion Tool Components are available to Scorpion Vision Partners and users holding a Scorpion Vision SDK Licence.

%d bloggers like this: