A Study of 3D Point Cloud Feature Extraction Method Based on Multi-Feature Fusion

To address the sparsity and irregularity of 3D point cloud data and enhance feature extraction through multi-feature fusion, this paper proposes the Multi-Feature Fusion Point Cloud Network (MFF-PCN). The network computes hybrid features from each point's spatial coordinates and geometric attributes using a Local Spatial Geometry Information Encoding Module and an FPFH-based Feature Extraction Module. The local encoding module captures geometric relationships by encoding self-coordinates, relative coordinates, and point-pair features (PPF), while the FPFH module further refines local features with Simplified Point Feature Histograms (SPFH) and weighted neighbor combinations. Experiments show a classification accuracy of 91.3% and overall accuracy of 93.3% on the ModelNet40 dataset.