Urban tree species identification based on UAV hyperspectral and LiDAR data

Accurate classification of urban tree species is essential for the improvement of urban environment and rational distribution of urban space. However, urban tree species are distributed and diverse, and it is challenging to classify them accurately. We extracted spectral, textural and vertical structure information based on UAV hyperspectral and LiDAR data, and used a random forest classifier to classify 14 tree species in complex urban areas. The results showed that the accuracy of tree species classification was significantly improved by the addition of LiDAR data. The fusion of spectral, textural and vertical structural features was the most effective combined features, with an overall accuracy (OA) of 83.48% and a Kappa of 0.81, in which the addition of vertical structural features increased the OA by 7.77% and the Kappa by 0.09. This study provides an effective reference for accurately obtaining information on tree species in complex urban environments.