Three-dimensional point cloud model repairing method with shape perception

The invention discloses a method for repairing a three-dimensional point cloud model with shape perception. The method comprises the following steps: step 1, acquiring a data set; 2, preprocessing experimental data; 3, designing a neural network model structure; step 4, constructing a neural network model loss; 5, building an experimental environment and training a neural network model; and step 6, storing the optimal model and various parameters thereof. By adopting the technical scheme, the incomplete and incomplete point cloud model with non-uniform point distribution is used as input, and the complete point cloud model with uniform point distribution after completion and repair is output. 本发明公开了一种具有形状感知的三维点云模型修复方法，包括以下步骤：步骤一、获取数据集；步骤二、预处理实验数据；步骤三、设计神经网络模型结构；步骤四、构建神经网络模型损失；步骤五、搭建实验环境并训练神经网络模型；步骤六、存储最优模型及其各项参数。采用上述技术方案，将不完整的、残缺的、点分布不均匀的点云模型作为输入，输出补全修复后点分布均匀的完整点云模型。