My passions are software development, pictures and mountains
In different projects I gained experences about image recognition, processing point clouds derived from ALS and TLS as well as database development. The topic artificial neural networks ANN is fascinating me as much that I 've started busy oneselfe with it.
A short example:
For calculation of Digital Elevation Model DEM you have to decide between local or global algortihms.
A simple local algortihm is Inverse Distance Weighting IDV algorithm. Simplyfy a elevation point is calculated by the sum of wights (the inverse distances) multiplied by hight. This sum devided by the sum of weights results the elevation of grid point.
A more complex approch is to use the Radial Basis Funktion RBF. In this strategy a elevation vector will be transformed into a high dimensional elevation polygon KRBF (xi,xj) = e-0,5||xi,xj||² . More details soon about it.
To aviod errors, privious filtering of distortions is neccesarry. Additionaly it has to be removed vegitation and buildings from the Digital Elevation Model DEM to get the Digital Terrain Model DTM.
After creating of DTM it is possible to compare it with a different point cloud of the same area. Thereby the nearest neighborhood or bilinear interpolation algrotihm is used for calculating a ground point. Cloud point minus ground point results the hight differnece. From all measures points of a gabled roof you get two planes. The intersection line of this planes is equal the calculated ridge line or ridge hight of the roof.
Druckversion 