Location: Range Management ResearchTitle: Multispectral remote sensing from unmanned aircraft: development of workflows and comparison with WorldView-2 data) Author
|Rango, Albert - Al|
Submitted to: American Society for Photogrammetry and Remote Sensing Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 9/5/2011
Publication Date: 3/19/2012
Citation: Laliberte, A.S., Rango, A. 2012. Multispectral remote sensing from unmanned aircraft: development of workflows and comparison with WorldView-2 data. American Society for Photogrammetry and Remote Sensing Proceedings. CD ROM. Interpretive Summary:
Technical Abstract: Unmanned aircraft systems (UAS) have seen increasing use in remote sensing of natural resources in recent years. Relatively low operation costs, ability to rapidly revisit the same location, and very high resolution imagery offer new opportunities for remote sensing applications and comparison with satellite imagery. Most digital cameras mounted on UAS deliver high spatial resolution imagery, but the low radiometric and spectral resolution has limits for vegetation mapping. We integrated a lightweight multispectral camera (Mini MCA 6, Tetracam, Inc.) into a BAT 3 UAS. The 6-band camera acquires 10-bit data from the visible to the near infrared spectrum, controlled by interchangeable filters, here designed to acquire data with band centers at 450, 550, 650, 720, 750, and 850 nm. Imagery was acquired at a flying height of 210 m, with a nominal GSD of 14 cm, over arid shrubland in southern New Mexico. We developed a custom approach for band-to-band registration and evaluated registration errors. Batch processing algorithms were developed for conversion of raw to tif file format, band-to-band registration, bit conversion, band stacking, and radiometric correction for convenient and fast processing of hundreds of UAS images. An empirical line method was used to fit digital numbers to field measured reflectance spectra. Spectral information derived from the orthorectified UAS mosaic was compared with data derived from a WorldView-2 image corrected atmospherically using ATCOR. We obtained good agreement (R2=0.82) for vegetation and soil reflectance values for the two data sources, especially in the red and near infrared bands. This demonstrates that high resolution multispectral UAS imagery shows great promise as ground truth for WorldView-2 data. An object-based classification of the UAS mosaic showed good spectral separation between arid land shrubs and grasses. We discuss challenges encountered in the multispectral image processing workflow and potential opportunities for multispectral, multi-scale image fusion.