DFG-PCN: Point Cloud Completion with Degree-Flexible Point Graph

2025-09-28

AI Summary

Point cloud completion suffers from geometric incompleteness due to occlusion and sensor limitations, especially struggling with fine-grained structures and discontinuous regions. To address this, we propose a degree-adaptive point graph-based completion framework. First, we design a node degree adaptation mechanism jointly driven by curvature and feature variation, enabling explicit perception of fine-grained geometric complexity. Second, we introduce Manhattan distance-weighted edge aggregation and a geometry-aware graph fusion module to strengthen collaborative local–global feature modeling. Third, multi-scale feature fusion is employed to enhance detail recovery fidelity. Evaluated on ShapeNet and other standard benchmarks, our method achieves significant improvements over state-of-the-art approaches in key metrics—including Chamfer Distance (CD) and F-Score—while qualitative results demonstrate robust reconstruction capability for complex, intricate geometries.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Point cloud completion is a vital task focused on reconstructing complete point clouds and addressing the incompleteness caused by occlusion and limited sensor resolution. Traditional methods relying on fixed local region partitioning, such as k-nearest neighbors, which fail to account for the highly uneven distribution of geometric complexity across different regions of a shape. This limitation leads to inefficient representation and suboptimal reconstruction, especially in areas with fine-grained details or structural discontinuities. This paper proposes a point cloud completion framework called Degree-Flexible Point Graph Completion Network (DFG-PCN). It adaptively assigns node degrees using a detail-aware metric that combines feature variation and curvature, focusing on structurally important regions. We further introduce a geometry-aware graph integration module that uses Manhattan distance for edge aggregation and detail-guided fusion of local and global features to enhance representation. Extensive experiments on multiple benchmark datasets demonstrate that our method consistently outperforms state-of-the-art approaches.

Problem

• Completes partial point clouds from occlusion and sensor limitations
• Addresses uneven geometric complexity distribution in shape regions
• Improves reconstruction of fine details and structural discontinuities

Innovation

• Adaptively assigns node degrees using detail-aware metric
• Uses Manhattan distance for geometry-aware edge aggregation
• Integrates local and global features with detail-guided fusion