形状特徴の重ね合わせによる個体別手指モデルの生成

概要

論文の詳細を見る
In this paper, we propose a method to synthesize subject-specific three dimensional finger models including skin and bones from geometric database and representative dimensions that can be measured by a caliper. The geometric database necessary for our method was developed by analyzing principal feature of the individual differences by using principal component analysis. The proposed method synthesizes subject-specific models by blending these principal features. The weight for each feature was determined by numerical optimization so that the synthesized subject-specific model reproduces the representative dimensions of the actual subject. We applied our method to fifty fingers. The average of the maximum error was 1.0mm, while the average error was 0.2mm. For more than 70% of the subjects, the bone as well as skin geometry were synthesized with acceptable error range from the representative dimensions of surface geometry. As for the remaining 30%, there was significant error in bone length. This method has distinct advantage over the conventional method in which time consuming manual segmentation for medical images is necessary, since we can generate subject-specific finger models in 5 to 10 minutes automatic computation.