Comparison of airborne and satellite high spatial resolution data for the identification of individual trees with local maxima filtering

Abstract

High spatial resolution airborne remotely sensed data have been considered a test bed for the utility of future satellite sensors. Techniques developed on airborne data are now being applied to high spatial resolution imagery collected from remote sensing satellites. In this Letter we compare the results of local maxima (LM) filtering for the identification of individual trees on a 1 m spatial resolution airborne Multi-detector Electro-optical Imaging Sensor II (MEIS II) image and a 1 m IKONOS image. With a relatively large spatial extent, comparative ease of acquisition, and radiometric consistency across the imagery, IKONOS 1 m spatial resolution data have potential utility for forestry applications. However, the results of the LM filtering indicate that although the IKONOS data accurately identify 85% of individual trees in the study area, the commission error is large (51%) and this error may be problematic for certain applications. This is compared to an overall accuracy of 67% for the MEIS II with a commission error of 22%. Further work in developing LM techniques for IKONOS data is required. These methods may be useful to forest stewards, who increasingly seek spatially explicit information on individual trees to serve as the foundation for more accurate modelling of forest structure and dynamics.

Acknowledgments

Paul Boudewyn, David Hill and David Seemann of the Canadian Forest Service (Pacific Forestry Centre) are thanked for collecting quality field data.