Urban Tree Species Classification Combining Spaceborne LiDAR and Multispectral Imagery

Abstract:

The urban tree species are an important factor affecting the ability of carbon sequestration by urban forest and the maintenance of ecosystem stability. However， due to the wide spatial distribution and complex environment of urban trees， there is a lack of tree species classification models applicable to cities. In this paper， the spaceborne LiDAR is introduced into tree species classification. Considering the vegetation canopy structure， horizontal spectra and spatial environment characteristics， the optimal feature set is constructed by quantitatively measuring the contribution of each parameter through feature space analysis. Finally， an urban tree species classification model combining spaceborne LiDAR and optical images is established using support vector machine （SVM） algorithm. Four representative regions in Shanghai are selected for validation， and the results show that the proposed fusion model has a high accuracy with the Kappa coefficient reaching 0.82 and the overall classification accuracy of 87.04%. The spaceborne LiDAR plays an important role in the urban tree species classification， and its retrieved 3D structural variables of vegetation together with spatial environmental characteristics play a major contribution to urban tree species classification.