遺伝的アルゴリズムに基づく教師つきテクスチャ分類用トレーニングエリアの自動選定   :   Automatic selection of reasonable training areas for supervised texture classification based on the genetic algorithm 

作成者 内山, 克史, 梶原, 康司, 張, 熙, 吉川, 敏則, 奥村, 浩
日本十進分類法 (NDC) 512.75, 007
内容 研究概要:An automatic selection method of reasonable training areas for supervised texture classification is proposed. This method is based on both the texture representation method using rotation invariant moments and the genetic algorithm (GA). In the proposed method, first, sets of candidates for each texture category are roughly given. The chromosomes for the GA are formed by the indexed candidates and their extent. The fitness function for the GA is obtained by the mixed texture model and quantified pure textures. In this paper, summary of the texture representation method and comparison of texture discrimination ability with the conventional Haralick's method are described. Furthermore, the details of the proposed automatic selection method are presented, and some simulation results are also shown.
コンテンツの種類 研究報告書 Research Paper
DCMI資源タイプ text
ファイル形式 application/pdf
掲載誌情報 第2回CEReS環境リモートセンシングシンポジウム論文集 page.69-76 (1999)
言語 日本語

Total Access Count:

591 times.

Related Materials in