Remotely piloted aircraft imagery for automatic tree counting in forest restoration areas: a case study in the Amazon

Abstract

Throughout the world, restoration of degraded areas (RDA) is not only a global but also a local challenge. In this context, the Brazilian government committed itself to restore 12 million hectares of forests by 2030. RDA monitoring customarily depends on extensive fieldwork to collect data on all individuals planted. As remotely piloted aircrafts (RPAs) can reduce costs and time of fieldwork activities, studying this technology is therefore timely given. A crucial metric for RDA is the number of trees established in the area. Methods using RPAs on automatic tree counting showed good accuracy using algorithms based on the canopy height model (CHM), which is the difference between a digital surface model (DSM) and a digital terrain model (DTM). However, obtaining a DTM demands an extra computational processing step and may require field control points or manually delimiting objects on the surface. The study presented here proposes and evaluates a semi-automated methodology for counting trees directly on DSM in RDAs in the Amazon using RPA coupled with a red–green–blue standard photographic sensor. The DSM method obtained good overall accuracy and F-score indexes, superior to the CHM method for all study areas even when overall accuracy was low for both methods.