Automatic generation of plant distributions for existing and future natural environments using spatial data

Abstract

This research proposes an algorithm for the generation of realistic plant distribution for both existing and future areas using spatial data. The main motivation for this research is to be able to generate realistic plant positions for the 3D visualization of existing and future natural environments. Current techniques for determining plant positions are limited. Plant detection techniques from the remote sensing domain are only able to detect positions for large plants using high-resolution imagery and LiDAR data. Often, the plant type of each detected position is not known. In addition, imagery and LiDAR data are not available for future areas. This research demonstrates that spatial data for future can be generated by using dynamic ecological models. These models can produce height, biomass, and coverage maps for future areas. The proposed algorithm must be able to translate data from existing areas as well as data produces by ecological models to a realistic plant distribution. A realistic plant distribution can contain plant positions for small and large plants with their corresponding plant type. To be able to realize this, an algorithm is proposed that integrates concepts of procedural generation and ecological modeling. Different spatial datasets can be provided as input and are analyzed to obtain information of where certain plant types can or cannot be placed in the target area. The algorithm was tested on both an existing area and a future area generated by an ecological model. The results were validated with statistical and expert validation methods. The statistical validation showed that algorithm is able to map the spatial data input correctly to a plant distribution. The expert validation showed that the generated plant distributions were in general judged realistic and that most issues in each plant distribution were because of missing data input or due to low data quality. In the future, the algorithm could be used for different applications and research. One possibility is to improve the current plant detection technique by providing additional information about the positions of small plants and by providing information about the plant type for each point with this algorithm.