High-resolution mapping of upland swamp vegetation using an unmanned aerial vehicle-hyperspectral system
Bikram Pratap Banerjee,a Simit Ravalb,* and Patrick Joseph Cullenc aAustralian Centre for Sustainable Mining Practices, School of Mining Engineering, University of New South Wales, Sydney, Australia NSW 2052 bAustralian Centre for Sustainable Mining Practices, School of Mining Engineering, University of New South Wales, Sydney, Australia NSW 2052. E-mail: [email protected] cSchool of Chemical Engineering, University of New South Wales, Sydney, Australia NSW 2052
Mapping of vegetation species and communities in sensitive ecosystems is essential for identification and management of anthropogenic impacts. Unmanned aerial vehicle (UAV)-hyperspectral systems are among the latest technologies in remote sensing that hold a potential for obtaining unprecedented quality of remote sensing data for vegetation mapping and health status monitoring applications. In this study, high-resolution (1–1.5 cm) spectral imaging data (15 bands) from a tunable spectrometer is used to map five species of vegetation in a complex upland swamp environment. The overall accuracy of classification was found to be 88.9% with a kappa coefficient of 0.83. Three classes (bare earth, sedgeland grass and black sheoak) have achieved higher accuracy (above 78%) and one class (bracken fern) has lower accuracy (58%). UAV-hyperspectral technology is, therefore, an effective tool to identify and map sensitive swamp vegetation. The technology can be potentially applied to determine the health status of the species.