Application of unmanned aerial vehicle for detection and classification of benthic macroalgal bloom

Abstract

Green tide mainly caused by Ulva sp. have been frequently observed along the coast of Korea every year. In particular, Jeju Island is main tourist site in the Korea, spending a lot of cost to remove green algae every year. Although satellite images provide a clear picture of the distribution of these macroalgal blooms, there were many difficulties in detecting small scale and/or submerged algae. This research attempts to identify the distribution and areal extent of the benthic green tide using unmanned aerial vehicles (UAVs). In mid-May 2019, the study was conducted on the southeast coast of Jeju Island, where massive benthic Ulva blooms occurred. The UAV images were taken at an altitude of 250 m using the DJI’s Mavic 2 Pro with optical camera. Individual images taken by UAV were produced as orthophoto through geometric correction, and the green algae were classified using the Maximum Likelihood, Mahalanobis Distance and Minimum Distance classification algorithms. Overall accuracy of Maximum Likelihood, Mahalanobis Distance and Minimum Distance classification method showed 97.6, 88.0 and 91.1%, respectively. In Kappa value as well as in overall accuracy of error matrix, the Maximum Likelihood method showed the highest accuracy. Despite some green algae were submerged, this study confirmed that benthic green algae could be classified with relatively high precision through an UAV images analysis.