Using airborne hyperspectral remote sensing (CASI) for operational forestry applications

摘要

Over the past four years, ITRES has implemented a number of operational programs determining various forest parameters for different stand types in Canadian boreal forests and west coast forests using the CASI hyperspectral scanner. Illumination conditions included totally overcast, sunny and partly cloudy conditions, some with low sun angles and non-optimal flight directions imposed by topography. The approach was to optimize the image collection efficiency and develop more robust interpretation algorithms able to handle the illumination variations. The initial focus was assessing the stem densities in regeneration areas (typical tree heights of 1 to 2 meters). CASI imagery, typically at 60cm pixel size, was combined in an automated fashion with digital elevation models (DEM's) to generate georeferenced mosaics incorporating multiple flight lines. The counting algorithms were expanded to permit automated counting of a wide range of tree sizes without pre-stratifications into polygons. Georeferenced vectors defining polygons of interest were then imported and relevant statistics generated as attributes for export to GIS. Similarly georeferenced vector files defining roads, rivers and other attributes could be Superimposed on the CASI imagery for output maps. Analysis products were expanded to include mapping of crown sizes, forest openings, candidate areas for fill-planting and regeneration areas having intrusions of less desirable species. All of these products were generated using automated processing techniques without the need for subjective interpretations by an interpreter. This paper will provide an overview of image interpretation methodologies, with emphasis on new developments in stem counting methods. Examples are from programs with several different forestry companies.