Pre-processing Techniques in Individual Tree Crown Information Extraction
As the extraction of individual tree information from high spatial resolution satellite and aerial images is getting more operational, it becomes increasingly important to examine some of the pre-processing techniques that will facilitate the semi-automatic analysis of large areas. For example, because of its main assumption that tree crowns are surrounded by shade, the Individual Tree Crown (ITC) software suite operates well in medium to dense forested areas, but, by-itself, breakdowns in more open areas.
This paper will examined techniques such as: separating forests from other vegetation and from non-vegetated areas; matching image resolution and/or smoothness to improve crown extraction; using medium density Lidar to match analysis tools to forest situations(i.e., mature, regenerating, open forests); applying BRDF corrections to aerial strips to permit proper spectral classifications; and introducing neighbourhood texture to separate different maturity levels within species.
Results from a variety of simple techniques, sensors and situations will be presented and discussed.