Is there a use of machine learning algorithm (e.g. Self-Organizing Maps (SOM), Graph Convolution Neural Networks (GCNN)) for vector spatial data using topographic maps? e.g. seperating/clustering linear features (transportation or hydrography) such as
- highways vs. railroads
- ditches vs main rivers
Merve Keskin if your input data is a continuous vector field (at each [x,y,z] you have a vector) then you could take a look at vessel detection[1] in medical imaging, that solves similar problems (detect connected 'flow' networks)
You can also use the classical Frangi filters[2] (on that vectorfield).
If your data is point based (e.g. Lidar), remote sensing has had good results last few (5-6) years with deep learning[3]
More 'exotic' approaches also exist, e.g. if you have a graph (sparser than 3d points) you can use network analysis/flow networks/graphlets etc..
[1] https://www.sciencedirect.com/science/article/abs/pii/S1361841519300982
[2]https://scikit-image.org/docs/0.14.x/auto_examples/filters/plot_frangi.html
[3]https://ieeexplore.ieee.org/abstract/document/8113128/
Dear Ben Cardoen , thank you very much for your answer and the references you shared. I guess [1] and [2] are very relevant with what we are trying to do.
Dear Merve Keskin
I recommend to read this article first:
Article Advances of Four Machine Learning Methods for Spatial Data H...
Hope this helps
Kind Regards
Qamar Ul Islam
Dear Qamar Ul Islam thank you very much but from my quick reading, it is for remotely sensed data (SPOT 7), we are more interested in vector data classification algorithms.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs