Predicting forest age classes from high spatial resolution remotely sensed imagery using Voronoi polygon aggregation

Efficient identification of forest age is useful for forest management and ecological applications. Here we propose a user-assisted method for determining forest age using high spatial resolution remotely sensed imagery. This method requires individual trees to be extracted from imagery and represented as points. We use a local maximum filter to generate points that are converted to Voronoi polygons. Properties of the Voronoi polygons are correlated with forest age and used to aggregate points (trees) into areas (stands) based on three forest age classes. Accuracy of the aggregation ranges from approximately 68% to 78% and identification of the mature class is more consistent and accurate than the younger classes.