Automatic Individual Tree Detection from Combination of Aerial Imagery, LiDAR and Environment Context.

Geographic Information Systems (GIS) allow analysis based on geo-referenced data. Currently only simple geo-referenced information is available, such as road networks or types of terrain, but there are other geo-referenced data that would be very useful to facilitate decision-making. These data are not collected as they are very hard to generate manually, but remote sensing data and artificial intelligence can be used to accomplish it. This work aims to develop an automatic framework for the extraction of geo-referenced trees, through the union Light Detection and Ranging (LiDAR) point clouds, aerial imagery, and existing GIS environment context. The results of the process are satisfactory, improving in some several areas the LiDAR-based detections using only imagery. However, issues such as false positives need to be corrected in the future. Merging both data sources would allow better results to be achieved.