風化の状態に非依存なテクスチャ特徴を用いた岩種の分類

概要

論文の詳細を見る
In this paper, rocks which contain some degrees of weathering are classified by a rock type using the support vector machine based on feature parameters of a boring core sample image. The feature parameters were determined based on visual features of the rocks. From a histogram of the image, the median value and the inter quartile range were used as feature parameters. From a co-occurrence matrix of the image, “angular second moment”, “correlation” and “information measures of correlation1 and 2” were used as features. The classification considered nonlinear separation by the support vector machine. A percentage of correct answers were about 71-86% on a test by a cross validation. In conclusion, it is shown that I1I2I3 color space and textural features are suitable for describing the rock type, and the rock type can be classified using the support vector machine with the feature parameters of color vectors and textural parameters.
社団法人電気学会の論文
2009-08-01