Effects of tree crown delineation in individual tree species classification with hyperspectral and LiDAR data

The effect of tree crown delineation on tree species classification using hyperspectral and LiDAR data was analyzed in this study. We considered a manual and an automatic crown extraction procedure. The results showed that crown delineation influenced the classification results. Using manual segments, a kappa accuracy of 0.86 was reached, while using only LiDAR segments it decreased to 0.74. Crowns automatically extracted from LiDAR data, in general, contained shadowed pixels that disturbed the classification process. The results also showed that a proper selection of the pixels inside the segmented crowns improved the classification results