Enhanced 3D Sub-Canopy Mapping via Airborne/Spaceborne Full-Waveform LiDAR

The Digital Imaging and Remote Sensing Image Generation (DIRSIG) software can generate geometrically and radiometrically accurate light detection and ranging (LiDAR) data, producing ground 3D (structural) truth data that would be nearly impossible to collect in complex forest environments. DIRSIG was leveraged to understand the nuances of a simulated forest, specifically Harvard Forest in Petersham, MA (Figure 1). This simulated forest included various materials such as bark, leaves, soil, and miscellaneous objects. LiDAR systems (airborne and spaceborne), which are useful for penetrating sub-canopy layers, were configured to collect LiDAR data. The resulting data sets provide valuable insights into forest health and sub-canopy intricacies, with applications in target detection, environmental monitoring, and forest management.

Two tools for automatic scene generation were developed. The first one, leaf cloud generator, generates scenes with a collection of leaves (Figure 1) to simulate the forest canopy layer with different features. This tool is used to assess the influence of varying forest characteristics on the LiDAR penetration capabilities. The second tool called forest bottom generator automates the process of producing diverse scenes of the forest floor with man-made objects (MMO) (Figure 2). The scenes generated by the second tool will serve as training data for the models trained to detect MMO without any disturbances coming from the higher forest layers. We plan to evaluate the models with different levels of signal sparsity. The evaluation under difficult conditions (low point density on the ground) will be complemented by the result of the analysis of the density of the signal reaching the bottom layers of the forest performed with the use of the first tool.

Figure 1: A scene generated by the leaf cloud generator.

Figure 2: A scene generated by the forest bottom generator.