关键词: 点云滤波, 局部特征, 动态图边卷积, 特征融合, 注意力机制

Abstract: In order to solve the problem that the existing filtering methods have feature ambiguity in dealing with point cloud models with high complexity geometric structure, which leads to poor filtering effect, this paper proposes a point cloud feature regularization fusion network (PFRNet) based on adaptive feature extraction and feature fusion theory. Firstly, the feature information between different neighborhoods was learned by the adaptive spatial feature extractor, so as to capture the local neighborhood features of different dimensions. Then, the global bilinear response is introduced from the local information of the point cloud through the local feature regularization fusion, and it is regularized and fused to weaken the common features of the point cloud and enhance the sharp features. Finally, the correlation attention decoder is used to enhance the connection between different neighborhoods in the decoding process, improve the global perception ability of the model, and better extract local geometric features. Experimental results show that compared with Pointfilter, PFRNet has 7.45% and 4.99% reduction in Chamfering Distance (CD) and Mean Square Error (MSE). The visualization results show that the proposed method can generate a point cloud model closer to the real one than other methods.

Key words: point cloud filtering, local feature, dynamic graph edge convolution, feature fusion, attention mechanism