Three-dimensional Point Cloud Object Recognition Method Based On Deep Learning

Three-dimensional point cloud object recognition refers to the analysis and processing of three-dimensional point cloud data to identify the types of objects contained therein.This is a hot research direction in the field of computer vision and is widely used in autonomous driving,robot perception,and other fields.However,existing methods suffer from issues such as insufficient feature utilization,poor accuracy,and low robustness.To address these problems,this paper proposes two deep learning-based methods for three-dimensional point cloud object recognition.The first method is the Fusion Attention Dynamic Graph Convolutional Neural Network(FA-DGCNN),which uses a feature extraction module based on fusion attention to obtain more discriminative feature maps.In addition,the model incorporates the idea of dense connections,where the input to each feature extraction module is directly connected to all previous output features,thereby improving feature reuse and model performance.The second method is the Multi-Scale Fusion Convolutional Neural Network(MSFCNN),which uses farthest point sampling to obtain point cloud data with different numbers of points.The model uses a single-scale feature extraction module to obtain feature information of different scales and a multi-scale feature fusion module to adaptively fuse feature information of different scales,thereby improving model accuracy and robustness.In experiments,the two methods achieved 93.9%and 93.8%accuracy on the ModelNet40 dataset,providing effective solutions for the problem of three-dimensional point cloud object recognition.