基于视觉形状与相关反馈的三维模型检索

摘要

The surface light source descriptor based on wavelet moment and improved Contour descriptor based on Gaussian potential function are proposed,and a 3D model search strategy with two stages is built combining the Adaboost relevant feedback. First,optimal bounding box of the model is got ,then Two-phase search is implemented ：the first stage search is locating the search range quickly based on Contour descriptor that is based on Gaussian potential function; the second stage search is accurate search mixing the two kinds of descriptor above. Last, the similarity of models is evaluated effectively combining the Adaboost relevant feedback. The re sult of experiment shows that this method is efficient and accurate.%提出一种基于小波矩的面光源描绘子和改进的基于高斯势函数的轮廓描绘子,并在此基础上结合Adaboost相关反馈,构建一种两阶段的三维模型检索策略.首先获得模型的最优包围盒;然后进行两阶段检索：第一阶段是基于高斯势函数的轮廓描绘子的快速定位检索范围,第二阶段是融合以上两种描绘子的精确检索.最后结合Adaboost相关反馈,从而有效地评估模型之间的相似度.实验结果表明：该方法具有较高检索效率与精度.