Unmixing Analysis of Close-Range Hyperspectral Imagery of Vegetation

Close-range hyperspectral imaging is a valuable but often underutilized tool for rapid, non-destructive and automated assessment of vegetation functional dynamics in terms of both structure and physiology. During the 2017 summer growing season, several hyperspectral images were collected at close range over a variety of vegetation plots consisting of vascular and non-vascular species and across a variable soil moisture gradient, which are associated with the Arctic Observing Network - International Tundra Experiment (AON-ITEX) in Utqiaġvik (formerly Barrow, Alaska). Hyperspectral images were collected with both visible near infrared (VNIR- 400-1000 nm) and short-wave infrared (SWIR- 900-1700 nm) hyperspectral imaging spectrometers developed by Surface Optics Corporation (SOC). Coincidental field spectra was also collected using a full range (300-2500 nm) HR-1024i field spectrometer developed by Spectra Vista Corporation (SVC).

Unmixing has been utilized traditionally to analyze satellite and airborne hyperspectral imagery of low to moderate spatial resolution. Here we present results on how unmixing algorithms can be utilized to extract unique spectral signatures in very high resolution imager such as the plot image and extract their spatial distribution. We study how the proposed approach handles illumination and spectral variability of the materials in the scene which are important aspects in this type of imagery. We present the potential of this approach to extract plant community structure to support image exploitation for ecological research efforts such as AON-ITEX.