Object extraction based on 3d-segmentation of LiDAR data by combining mean shift with normalized cuts: Two examples from urban areas

Wei Yao, Stefan Hinz, Uwe Stilla

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

26 Citations (Scopus)


In this work, we have looked into the problem of urban analysis using airborne LiDAR data based on the strategy of classification by segmentation. Segmentation is a key and hard step in the processing of 3D point clouds, which is not perfectly solved in view of different applications. A new 3d segmentation method incorporating the advantages of nonparametric and spectral graph clustering is presented here to facilitate the task of object extraction in urban areas. This integrated method features local detection of arbitrary modes and globally optimized organization of segments concurrently, thereby making it particularly appropriate for partitioning raw airborne LiDAR data of urban areas into segments approximating semantic entities. Two examples in urban areas - flyover and vehicle are chosen as interest objects to be extracted by a classification-based step. The approach has been tested on LiDAR data of dense urban areas, and the results that are obtained have been compared with manual counts and showed us the efficiency and reliability of the strategy.
Original languageEnglish
Title of host publication2009 Joint Urban Remote Sensing Event
Publication statusPublished - 26 Oct 2009
Externally publishedYes
Event2009 Joint Urban Remote Sensing Event - Shanghai, China
Duration: 20 May 200922 May 2009


Conference2009 Joint Urban Remote Sensing Event

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Networks and Communications
  • Ecology
  • Nature and Landscape Conservation

Cite this