Today, tree distribution maps can only be generated manually in a very time-consuming process, and real-time microclimate mapping of a large 3D volume under tree canopy is not possible. A prototype small quadrotor unmanned aerial vehicle (UAV) system was developed that is able to maneuver in cluttered environments like forests, and under tree canopy to map tree distributions from 3D point clouds gathered from an onboard stereo vision system. The UAV uses a small onboard sensor board to record micro-climate parameters during flight.
Based on an existing micro air vehicle system that executes an autonomous navigation framework to maneuver in GPS-denied environments, close to the ground and in highly cluttered environments, a data processing framework was developed to analyze imagery from an onboard, forward-looking stereovision system to calculate tree trunk distribution maps. This includes estimating the forest floor profile, detecting tree trunks in the images, and calculating the diameter of tree trunks at a certain height above the ground. While data processing is intended to run onboard the vehicle, it is currently performed offline from collected images, which are manually labeled to compare tree diameter and position estimations with ground truth. The framework was tested on data from an example forest of live oak.
Applications for this work include rapid assessment of biomass for commercial forestry and carbon offsets, and science data collection such as structure and microclimate that are currently not feasible at this scale. Additionally, mapping below-canopy microclimate within forest canopies is an essential aspect of carbon cycle model validation, because it has the potential to falsify parameterizations that are currently poorly quantified. Microclimate sensing of radiative fluxes also has potential for improved remote sensing retrievals under complex canopies. NASA’s terrestrial modeling efforts in carbon and water cycles can be significantly improved by the proposed spatial data collection of microclimate. This approach also expands carbon inventorying and microclimate mapping into environments that are traditionally very difficult to map (e.g. mangrove mapping).