Learning-based multi-view stereo (MVS) is gaining prominence as a method for 3D reconstruction. However, existing methods in the process of feature learning fail to focus on the structural information implied in the scene. This oversight prevents the network from perceiving the geometric properties of the scene and weakens the generalizability of the network. Therefore, we propose a novel framework named Point Feature Relation Network for Multi-view Stereo (PFR-MVSNet), which is composed of a Dynamic Structure Perception (DSP) module, an Adaptive Feature Exploration (AFE) module, and a Point Transformer Block (PTB) module, to solve the problems caused by the oversight. The DSP module first augments the feature of the 3D point cloud from multi-view 2D features, then establishes spatial structure relations within local regions on the point cloud and guides the feature learning of points through the aggregated structure information. After the network has fully learned the intra-region structure features, the AFE module repartitions perception regions with similar features. The point features within the regions are further learned by the PTB module. We evaluate our method on three benchmark datasets: DTU, Tanks & Temples, and ETH3D. The experimental results show that our method achieves superior accuracy of 0.289 mm on the DTU dataset and exhibits more robust generalization on the Tanks & Temples and ETH3D datasets compared with other learning-based MVS methods.

The research on reassembling broken objects has many important applications, such as cultural relics restoration, medical surgery and solving puzzle. Because of the complicated surfaces of the fractured object pieces, it is not easy to extract salient features from them. It becomes even more difficult and very time-consuming to reassemble broken objects when the fragments are severely corroded or some of them are lost. In order to improve the accuracy and speed of 3D fragment reassembling, an effective and efficient fragment reassembling algorithm based on point clouds is proposed this article. This method first extracts keypoints and their concavity and convexity according to the symbolic projection distance of the point cloud, and then uses the local neighborhood information of the keypoints to construct a multi-scale covariance matrix descriptor. Furthermore, by calculating the similarity of the covariance matrix descriptors, the initial pairs of match points are obtained. Finally, the geometric constraints are gradually added to optimize the sampling so as to find good hypotheses as quickly as possible. By doing so, the search space is narrowed continuously in each iteration of the process to speed up the hypothesis test. We have conducted extensive experiments. The results show that the proposed method can fuse multiple features of the fragments effectively and achieve an outstanding matching effect on the defected fragments, and that the proposed method is faster than the existing methods in literature.

To overcome the limitations of inadequate local feature representation and the underutilization of global information in dynamic graph convolutions, we propose a network that combines attention mechanisms with dual graph convolutions. Firstly, we construct a static graph based on the dynamic graph using the K-nearest neighbors algorithm and geometric distances of point clouds. This integration of dynamic and static graphs forms a dual graph structure, compensating for the underutilization of geometric positional relationships in the dynamic graph. Next, edge convolutions are applied to extract edge features from the dual graph structure. To further enhance the capturing ability of local features, we employ attention pooling, which combines max pooling and average pooling operations. Secondly, we introduce channel attention modules and spatial self-attention modules to improve the representation ability of global features and enhance semantic segmentation accuracy in our network. Experimental results on the S3DIS dataset demonstrate that compared to dynamic graph convolution alone, our proposed approach effectively utilizes both semantic and geometric relationships between point clouds using dual graph convolutions while addressing limitations related to insufficient local feature extraction. The introduction of attention mechanisms helps mitigate underutilization issues with global information, resulting in significant improvements in model performance.

With the rapid development of point cloud processing technologies and the availability of a wide range of 3D capturing devices, a geometric object from the real world can be directly represented digitally as a dense and fine point cloud. Decomposing a 3D shape represented in point cloud into meaningful parts has very important practical implications in the fields of computer graphics, virtual reality and mixed reality. In this paper, a semantic-driven automated hybrid segmentation method is proposed for 3D point cloud shapes. Our method consists of three stages: semantic clustering, variational merging, and region remerging. In the first stage, a new feature of point cloud, called Local Concave-Convex Histogram, is introduced to first extract saddle regions complying with the semantic boundary feature. All other types of regions are then aggregated according to this extracted feature. This stage often leads to multiple over-segmentation convex regions, which are then remerged by a variational method established based on the narrow-band theory. Finally, in order to recombine the regions with the approximate shapes, order relation is introduced to improve the weighting forms in calculating the conventional Shape Diameter Function. We have conducted extensive experiments with the Princeton Dataset. The results show that the proposed algorithm outperforms the state-of-the-art algorithms in this area. We have also applied the proposed algorithm to process the point cloud data acquired directly from the real 3D objects. It achieves excellent results too. These results demonstrate that the method proposed in this paper is effective and universal.

As 3D technologies advance rapidly, 3D printing, 3D animation, and 3D Movie are springing out in different areas. It becomes a remarkable challenge to generate a large number of 3D models effectively and efficiently. This paper proposes a novel editing method based on the feature lines of images (i.e., image contour and principal axis) for generating new 3D models. Our method takes as input an existing 3D model (as the original model) and an image selected by the user (or a sketch hand-drawn by the users), and performs the model editing to generate a new 3D model. In particular, our method first takes as input an original 3D model and an image selected by the user. Second, the selected image is processed to produce the feature lines, i.e., contour and principal axis of the image. Third, the silhouette of the original model from a given view is acquired, and projected on a projection plane to produce a contour and principal axis of the model, which is the feature lines of the original model. Fourth, by comparing the feature lines of the image and the original model, the constraint conditions are established to control the editing of the 3D model. Finally, 3D model editing is conducted through the as-rigid-as-possible mesh deformation to produce a new 3D model with the appearance resembling the selected image. Furthermore, this paper proposes an energy function to guide the detailed model editing, and measure the similarity between the generated 3D model and the corresponding image. We have conducted extensive experiments to evaluate the proposed method. The results show that comparing with the existing editing methods in literature, the proposed model editing method is able to construct various types of 3D models more effectively and more efficiently.