GEOSCAN Search Results: Fastlink


TitleUnsupervised Landscape Unit Mapping Based on Multi-scale Analysis
DownloadDownloads (Preprint)
LicencePlease note the adoption of the Open Government Licence - Canada supersedes any previous licences.
AuthorBeauchemin, M; Pan, D; Fung, K B
SourceThird International Conference on Geospatial Information in Agriculture and Forestry, Denver, Colorado, 5-7 November; 2001., Open Access logo Open Access
Alt SeriesEarth Sciences Sector, Contribution Series 20043005
Mediapaper; on-line; digital
File formatpdf
Released2001 01 01
AbstractWe present the preliminary development of a segmentation algorithm for landscape unit mapping. A multi-scale approach is used (i) to establish local proportion of land cover types within different window sizes centered around each pixel and (ii) to determine the suitability of each window size to support pattern unit candidacy. A fusion process of the multi-scale proportion map is then carried out at each pixel position based on minimum acceptable size of pattern unit. An unsupervised segmentation algorithm is finally utilized to cluster the resulting fused map into homogeneous regions. Preliminary results are presented for a 10-class land cover map derived from a supervised classification of a 30 km x 30 km subarea of a Landsat TM scene.

Date modified: