Remote Sensing of Aquatic Invasives via Multi-Resolution Analysis of Time-Series Data: Fusing MODIS Vegetation Signatures and NOAA Buoy Signals
Lori
Mann
Bruce, Mississippi State University, bruce@ece.msstate.edu
(Presenting)
Abhinav
Mathur, Optech International Inc., m_abhinav@hotmail.com
John
Madsen, Mississippi State University, jmadsen@gri.msstate.edu
Multi-resolution analysis, namely discrete wavelet transforms, are used to denoise and extract features from MODIS time-series data for the detection of aquatic invasive species. In addition, wavelet analysis is used to correlate the MODIS time series data to NOAA buoy signals in order to account for tidal fluctuations along the coastline which can significantly affect the reflectance properties of the target vegetation. Using the MODIS imagery, temporal vegetation signatures, i.e. vegetation indices as functions of time, are generated. Due to challenges with the MODIS quality assurance maps, a significant level of noise is present in the temporal signatures. Several methods for denoising the signatures are investigated, including simple moving average and median filters, as well as wavelet denoising techniques. The authors also have developed a wavelet-based feature extraction method for quantifying shape of the oscillations in the temporal signatures. This poster provides an explanation of how NOAA buoy time-series data was correlated to the MODIS time-series data to account for tidal effects on the vegetation signatures, as well as a comparative analysis of the denoising methods and wavelet-based versus Fourier-based feature extraction techniques.