Evaluating composition methods of SPOT-VGT S1 data for land cover mapping of Southeast Asia
Various pixel mosaic techniques and a sub-image mosaic technique of the SPOT- VEGETATION sensor data are compared for the purpose of exploiting this quotidian database (SI) for vegetation cover mapping in continental Southeast Asia for insertion into the Global Land Cover 2000 project. One- and two- criterion pixel composting techniques for the SI database are compared based upon the following four criteria: degree of sensitivity to the atmosphere, texture variance, sensitivity to various vegetation types, and view zenith angle tendencies. These comparisons are extended to dry season composites created using the IO-day synthesis product (SI 0), and to a sub-image mosaic. The sub-image mosaic technique is rejected based on the automated atmospheric corrections applied generically to the SI database, and angular effects causing a visible seam between the sub-images. The most useful results for mapping purposes are therefore obtained from pixel-based composites of heavily masked images for cloud, cloud shadow, imaging defects and scan angle (hot-spot and specular). The best overall result is generated by the two-criterion maximum NDVI /minimum red reflectance. It is anticipated that results would be improved if scene-specific atmospheric characteristics could be accounted for, and subsequent accurate bi- directional effect correction applications could be made and atmosphere-specific optimised vegetation indices could be used as composting criteria.
Bibliographic Reference: An oral report given at: 23rd Canadian Symposium on R.S. and the 10th Congress of the Quebec R.S. Association. Organised by: Universite Laval. Held in: Quebec (CA), 20-24 August (2001)
Record Number: 200013782 / Last updated on: 2001-09-18
Original language: en
Available languages: en