Individual Tree Crown Segmentation in Aerial Forestry Images by Mean Shift Clustering and Graph-based Cluster Merging

Individual tree crown segmentation is frequently required in forest inventory, biomass measurement, change detection, tree species recognition, etc. It is almost impossible to do manual segmentation of huge forest by human. In this paper, we present an automatic method for individual tree crown segmentation in aerial forestry images. We first extract treetops using the method in (1). Next we apply mean shift clustering to group pixels into clusters having homogeneous properties. Then we build a cluster adjacency graph where clusters belonging to the same crown are merged. We tested our method on some forestry images and obtained good results.