Research on Partial Missing Reconstruction of 3D Point Cloud Model Based
on Point Fractal Network
Abstract
In order to solve the problem of partial loss of data information and
structure of 3D point cloud model due to subjective and objective
factors such as occlusion and noise, a partial deletion reconstruction
method of 3D point cloud model based on PF-Net was proposed, and a model
deletion reconstruction system was developed. Based on the deep learning
PF-Net network architecture, the Batch Normal layer and the Dropout
layer are introduced to normalize the original datasets in batches,
which further improves the reconstruction efficiency and accuracy of
some missing point cloud models. In this paper, 11 point cloud models
are selected to carry out reconstruction experiments with some missing
data information and structural features, and the experiments show that
the proposed method has higher reconstruction efficiency than the L-Gan
and PCN methods when the same dataset is used for training and testing.
In the eleven test categories, the average improvement of the
refactoring method in this paper is 12%-27%. The proposed method has a
significant effect in dealing with the partial deletion reconstruction
of small-scale models, and at the same time improves the efficiency and
accuracy of reconstruction, which has good application value.