Need help with point cloud segmentation.

[vincentmarkiet]

Hello everyone,

I have an airborne lidar point cloud data set of a forest strip. Within this forest pointcloud there's a corridor of hanging powerlines. The raw data only has xyz values.

What I want to do is to create an algorithm that automatically extracts the powerlines,gives me the coordinates and measures the length, and curvature of each powerline.

I have a fair idea of an approach:

1 1. classify the las dataset : classify ground points, non-ground-points
2. Get rid of ground points and start working on non-ground points
3 By using a histogram in the east-west plane I would be able to detect the corridor since there will be less points.
4 Filter out the rest and only use the points inside the corridor.
5 Repeat the last step in a vertical direction to find the powerlines.
6 Extracts the lines from the towers.

Anyone has some tips for me how to do this in CloudCompare?
And will it be possible to create a simple script in Cloudcompare or through Python to have this automatically done?

greetings,
Vincent

[daniel]

Wow, quite ambitious ;)

There's several way to segment the power lines. Maybe the best one would be to use the CANUPO classifier (qCanupo plugin: http://www.cloudcompare.org/doc/wiki/in ... uginCANUPO). Should be quite easy to train in order to detect directly the power lines (as their "dimensionality" should be very singular). If you need help don't hesitate to send me a sample point cloud (cloudcompare [at] danielgm.net).