サンゴ被度の自動検出に向けたテクスチャ解析

Texture analysis for automatic calculation of coral cover

松田 朝陽*; 奥田 雄大*; 高橋 悟*; 川端 邦明   ; 小笠原 敬*

Matsuda, Asahi*; Okuda, Yudai*; Takahashi, Satoru*; Kawabata, Kuniaki; Ogasawara, Kei*

本発表は、視覚的にサンゴ被度を自動検出するためのテクスチャ解析に関するものである。サンゴ礁において白化状態の検出は重要な観察ミッションの一つであり、本研究ではこの検出支援を行うための画像処理手法を開発している。サンゴ礁画像, 海底画像にテクスチャ解析を適用して分類を行った実験結果について述べる。

Research of monitoring of corals plays an important role in observing the natural environment on a global scale. Identifying and detecting corals amongst other objects, including marine organisms, from an image taken undersea is a difficult task. Moreover, there are many types of corals and they cannot be easily categorized automatically. Currently, texture analysis is being pursued that solves the problem of detecting and recognizing corals from pictures taken undersea; Haralick 14 conditional equations are used to calculate the texture characteristics information based on the gray level co-occurrence matrix of 256 gradations and also to perform liner discrimination analysis on the characteristics information. In this paper, we propose one algorithm of texture analysis for discrimination of undersea objects, in order to detect the corals.