Paper Title
IDENTIFICATION OF SEA-ICE FLOES AND FLOE SIZE DISTRIBUTIONS USING IMAGE PROCESSING
Abstract
An unmanned aerial vehicle was used as a mobile sensor platform to collect sea-ice features at Ny-Ålesund in early May 2011, and several image processing algorithms have been applied to samples of sea-ice images to extract useful information about sea ice. The sea-ice statistics given by the floe size distribution, being an important parameter for climate and wave- and structure-ice analysis, is challenging to calculate due to difficulties in ice floe identification; particularly snake algorithm is applied to solve this problem. To evolve the GVF snake algorithm automatically, an initialization based on the distance transform is proposed to detect individual ice floes, and the morphological cleaning is afterward applied to smoothen the shape of each identified ice floe. Based on the identification result, the image is separated into four different layers: ice floes, brash pieces, slush, and water.. The proposed algorithm implements Otsu Thresholding method, K-Means clustering algorithm, Distance transform, edge detection and morphology operation yields an acceptable identification result. A discussion on the methods and results concludes the paper.
Index Terms: Floe size distribution, ice floe identification, image processing, marginal ice zone, remote sensing.