Remote sensing of forest physical structure currently relies mainly on satellite data, aeroplanes, and above-canopy drones, but sensors on these above-canopy platforms have difficulty penetrating into deeper layers of forests, especially in dense evergreen tropical forests. Below-canopy drones can complement above-canopy surveys and provide more-holistic assessments of forest structure.
We teamed up with our colleague Feng Lin, formerly of the Department of Electrical and Computer Engineering at NUS and now of Peng Cheng Laboratory in China, to use data from an autonomous drone flight in parkland for estimating tree diameters. The drone used LiDAR sensors and simultaneous localisation and mapping (SLAM) to navigate the small area of parkland, and in post-processing we used the LiDAR and SLAM data as inputs to automated algorithms for detecting and measuring trees. The automated measurements of tree diameter were closely correlated with subsequent manual measurements (R2 = 0.92). The study has just been published in Remote Sensing.
This study is a step towards fully automated below-canopy forest assessment, although many challenges remain, including the development of software for autonomous navigation of real forests, which are typically more complex than parkland.