为解决不完整三维模型修复过程中孔洞形状不规则、现有方法曲面配准精度不高的问题, 提出一种能够有效保持孔洞边界过渡自然并恢复模型表面细节特征的模型修复方法。首先, 追踪所有1-邻域边和1-邻域三角形数量不相等的点来探测模型的孔洞边界。设计一种基于二维网格数量的算法确定不完整模型的匹配候选集, 同时根据双稀疏表示的三角网格顶点位置误差、边平滑误差和正交约束来预测最优匹配模型; 然后, 结合边界顶点的曲率、边界轮廓线折角的余弦值及相邻边界点的线段长度构造能有效表达不完整模型与碎块模型之间对齐关系的特征点描述子; 最后, 使用二阶伞算子来平滑模型的修补边界。实验结果表明, 模型修复时间节省19% ~ 26%, 模型修复误差平均下降35%。该方法打破了当前模型修复方法中修复裂缝大且难以实现模型表面细节的局限性, 可快速有效地实现破损模型的修复。

To solve the problems of irregular hole shape and low accuracy of surface registration in an incomplete 3D model-repair process, a model-repair method was proposed to effectively maintain the natural hole boundary and restore the model surface details. First, we traced all points with unequal numbers of one-neighborhood edges and one-neighborhood triangles to detect the hole boundaries of the model. A method based on 2D mesh numbers was proposed to determine the matching candidate set for the incomplete model, and the optimal matching model was predicted based on the vertex-position error, edge-transition error, and orthogonal constraint, which were represented by double sparstiy. Then, the curvature was combined, cosine value of the boundary contour angle, and length of the line segments of the adjacent boundary points centered on the same boundary vertex to construct the feature descriptors that can effectively express the alignment relationship between the incomplete and fragmented models. Finally, the second-order umbrella operator was used to make the repair boundary of the model smooth. The experimental results demonstrate that the repair time is reduced by 19%—26% and the repair error is reduced by 35%, on average. This method avoids the limitations of large cracks and poorly realized surface details of the current model-restoration methods and can quickly and effectively repair the damaged model.