Classification of urban tree species using hyperspectral imagery

Urban areas serve as humanity's principal habitat. Because of this, it is important to understand the biophysical components of the urban environment – including the urban forest. The goal of this study was to determine the potential to classify individual urban trees as a function of spectral features derived from airborne hyperspectral data. To determine this, 500 urban trees were identified (through fieldwork) in the built-up zone of Provo-Orem, Utah, USA. Visible and near infrared airborne hyperspectral imagery was collected over the same area. The 500 trees were identified on the images, and spectral features of each tree were extracted. Principal components, vegetation indices, band means, and band ratios were all used as features to discriminate between different tree species. The tree classification was 82% accurate when just the six principal components were used. Classification accuracy increased to 91.4% after combining vegetation indices, band mean values and band ratios.