Lee, Y.-S.; Park, S.-H.; Cho, Y.-H.; Lee, W.-J, Jung, H.-S.; Lee, M.-J., and Kim, S.-H., 2019. Classification of halophytes from airborne hyperspectral imagery in Ganghwa Island, Korea using multilayer perceptron artificial neural network. In: Jung, H.-S.; Lee, S.; Ryu, J.-H., and Cui, T. (eds.), Advances in Remote Sensing and Geoscience Information Systems of Coastal Environments. Journal of Coastal Research, Special Issue No. 90, pp. 243-250. Coconut Creek (Florida), ISSN 0749-0208.

Surveying methods have limitations when investigating the vegetation of mudflats, which are difficult to access due to tidal variations. The use of remote sensing data to explore inaccessible targets and regions is an effective approach to analyzing vegetation species. Recently, hyperspectral images, composed of several tens to several hundreds of spectral bands, have been effectively utilized to classify and analyze halophytes using spectral reflectance differences. In this study, halophytes were classified from an airborne hyperspectral image of the Dongmak beach using an artificial neural network (ANN) approach. Dongmak beach is located on Ganghwa Island, Korea, where mudflats are well developed. Since the hyperspectral image has many spectral bands, it was tried to reduce the number of bands. To do this, the statistical values of halophytes and mudflats from all of the hyperspectral bands were first calculated by using several methods, including parallelepiped, Mahalanobis distance, minimum distance, spectral information divergence, and a spectral angle mapper. The statistical images were then used to classify halophytes and mudflats by applying a multilayer perceptron artificial neural network (MLP-ANN). The achieved accuracy of the MLP-ANN classification was about 95.02 %. Some areas composed of halophytes were misclassified as mudflats. These results will reduce the time and cost of field investigations in large areas.