When: June 12, 12:00-14:00
Where: Hugo Priemus Room, OTB, Building 30, Campus TU Delft.
Presentation by Ravi Peters
Title: Ongoing research, smart thinning of LiDAR data with the Medial Axis Transform
Many existing toolchains for point cloud processing such as the construction of the 3D version of TOP10NL are resources hungry and are sometimes even limited by how much points fit main memory. Therefore, input point clouds are thinned, currently using simple techniques such as random or nth point filtering. With these methods, however, every point has an equal chance of removal. Yet, for many application not all points are equally important, which for instance means that a practitioner needs to set different thinning paramaters for rural areas than for urban areas.
I am currently investigating how to apply the Medial Axis Transform (MAT) for `smart` thinning, e.g. a thinning method that is adaptive to the geometric importance (i.e. local feature size) of points. Key to the successful application of this approach is the construction of a `clean` MAT, that is insensitive to the noise that is typically present in LiDAR data. In this presentation I will describe and demonstrate an number of techniques to deal with noise during the construction of the MAT. I will also elaborate a number of other interesting findings that may prove useful in a later stage of my research (i.e. MAT-based feature identification).
Short presentation: Download tool AHN2 (http://3dsm.bk.tudelft.nl/matahn)
As you may know the Ducht AHN2 dataset has recently been released to the public. While this is a great thing for Dutch point cloud enthousiasts, the current download procedure through our national geoportal is somewhat involved to say the least. In this presentation I will briefly introduce an alternative download tool developed at GISt, that allows the user to simply select a square area on a map and press ‘download’. On the long term we also plan to offer smartly thinned point clouds through this tool.