Remote sensing offers an economical tool to perform forest mapping and to meet forest manager needs. The main objective of this thesis is to compare the results of both supervised classification and linear spectral analysis methods of forest cover mapping from the same sector of Anticosti Island. Supervised classification is a binary method of mapping where the pixel can be associated with only one thematic class whereas linear spectral mixture analysis is based on the principle of continuity, meaning that a single pixel can belong to more that one information class.
To meet our objectives, the image used in this thesis was acquired on August 24 2001 by Landsat-7's ETM+ sensor. Radiometric calibration, atmospheric corrections and geometric corrections were applied to the image. Spectroradiometric data acquired by ASD (Analytical Spectral Devices) between 350 nm and 2500 nm were also used.
The results show that the supervised classification performed with parallelepiped algorithm combined with the maximum likelihood algorithm produced good results for forest mapping in a mixed environment. The results from this method can be used to perform multitemporal studies regarding forest cover observation on Anticosti Island i.e., disease infestation, forest fire, heavy browsing by deer and regeneration. However, the linear spectral mixture analysis method results were not as conclusive as expected because of the limited number of spectral bands that can be used, the spectral regions covered by those spectral bands, and the similarity between the spectral signature of the different forest species.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/26745 |
| Date | January 2004 |
| Creators | Pinard, Véronique |
| Publisher | University of Ottawa (Canada) |
| Source Sets | Université d’Ottawa |
| Language | French |
| Detected Language | English |
| Type | Thesis |
| Format | 162 p. |
Page generated in 0.0156 seconds