3D

By working with the full waveform in 3dimensional xyz, several added layers of information becomes available by the process of pulse reflection and echo. Because, by adding several layers of information within the DTM, such as nDSM, Echo Ratio, and surface roughness, it might be possible to recreate and derive a more accurate and complete pictures of potential information, and thus define best procedure and best extraction possibilities.

For 3dimensional building extraction (Höfle et al. 2009, 7):

“The outline extraction is per-formed sequentially for all candidate regions by extracting the laser points for each region. The applied method defines the building outline as the orthogonal projection of all roof points of a single building onto the xy-plane. It is initiated by a mean shift segmentation (MELZER 2007) resulting in a decomposition of the candidate point cloud into segments representing individual buildings or building parts. These segments are further subdivided into individual roof planes by means of a segmentation algorithm considering the local regression planes of the individual points. By merging roof segments belonging to one building, projecting their outline to the xy-plane, and finally regularizing this outline, the building outlines are derived. If none or too small roof planes are found, e.g. candidate region with solely vegetation, no building outline will be generated, which is an additional check (i.e. classification) of the candidate regions. Further details on roof outline generation and regularization are given in DORNINGER & PFEIFER (2008).”

(a) Comparison of detected candidate regions, modeled building outlines and digital cadastral map: (1)candidate region rejected in object-based classification, (2),(3)different representation of building in LiDAR data and cadastre, (4)oversegmentation of candidate regions; (b) shaded nDSM overlaid with final building outlines, and (c) CAD dataset of 3D building modeling output. (Höfle et al. 2009, 8)

(a) Comparison of detected candidate regions, modeled building outlines and digital cadastral map: (1)candidate region rejected in object-based classification, (2),(3)different representation of building in LiDAR data and cadastre, (4)oversegmentation of candidate regions; (b) shaded nDSM overlaid with final building outlines, and (c) CAD dataset of 3D building modeling output. (Höfle et al. 2009, 8) 

Leave a Reply

Your email address will not be published. Required fields are marked *