Boundary extraction (identification and tracing) from point cloud data

For any input point set the algorithms are able to not only identify boundary edges, but also trace.

The algorithms can:
1) extract (identify and trace) both outer and inner boundaries separately from the input point set,
2) work on any type of input point sets,
3) extract boundaries separately from each subset of a large and/or dense point set. The original input point set can be split (user defines how to split) and then (both inner and outer) boundaries can be extracted from the combined subset boundaries. This significantly reduces computational cost for a large and/or dense point set.
The algorithms should work for any point cloud data. The only input parameter is dmax, which is maximum point-to-point distance in the input point cloud. For the generated shapes (by shape_gen.m file) from the given sample data set, this value is in between 1.0 to 1.5 pixels. So, dmax = 1.3 pixels is used.

The algorithms have been extensively tested against numerous LIDAR point cloud data. For example, for the sample data in LIDAR_sample_01_adjusted.txt this second parameter is 0.2 (metre).

The algorithms are also capable of extracting multiple boundaries for more than one object in one single input data set. In that case, the distance between two objects should be at least 2 times the maximum point-to-point distance in the input point cloud.

Please refer the paper: M. Awrangjeb, "Using point cloud data to identify,
trace, and regularize the outlines of buildings" International Journal of Remote Sensing, Volume 37, Issue 3, February 2016, pages 551-579; Open access at: http://www.tandfonline.com/doi/pdf/10.1080/01431161.2015.1131868