Developing drone-based hyperspectral imagery for forest infestation detection

Purpose and goal:

We fulfilled the objective of the project to develop UAV-based hyperspectral imagery for the early detection of bark beetle infestations. We developed the methodology including data acquisition, spectral analysis, and proposed better models for infestation identification. We also fulfilled the objective of scientific mobility and research collaboration with the Finnish Geospatial Research Institute (FGI), by Dr. Langning Huo working at FGI for six months on the same research topic.

Expected results and effects:

We analyzed the spectral signatures of the healthy and infested trees and found that the crown spectrum in the green shoulder and red edge regions differed the most after the infestation. Based on this finding, we proposed new vegetation indices that can improve the identification of infestations at the early stage. We also simplified the vegetation indices adapted for multispectral sensors, which are cheaper and easier to collect data from than hyperspectral sensors.

Approach and implementation:

This project used state-of-the-art UAV-based hyperspectral sensors with high resolution and high frequency. The research design can help to sufficiently answer the scientific question of how early infestations can be identified. The hyperspectral images were processed and calibrated with high quality, supporting the spectral analysis. We concluded the most informative bands from hyperspectral sensors and proposed the implementation on multispectral sensors in practical use of forest monitoring.