Skip to Main Content
Several features extracted from airborne laser swath mapping (ALSM) data are examined to determine their effectiveness in separating buildings from trees across geographically and temporally diverse landscapes. These two classes are often spatially mixed in urban and suburban areas and can be quite difficult to separate based solely on geometric information due to the discrete sampling of ALSM. New median-based distance measures are used to quantify the separability of the classes using the different features. Information-based measures are also applied to the same data. For each of the test cases, it is possible to identify a common feature space in which the distance between the two classes is large. This distance information is an indication of the separability between classes and is therefore indicative of the potential success likely when trying to classify ALSM data. This analysis provides new insights into the richness of simple two-return ALSM data and to the spatial and temporal stability of ALSM features when discriminating between classes.