Point Cloud Classification Model Based on a Dual-Input Deep Network Framework

Ruifeng Zhai,Shuxu Guo,Junfeng Song,Xueyan Li,Yu Hou,Zhenxin Wang,Fengli Gao,Shuzhao Hou

doi:10.1109/access.2020.2981357

Abstract

The disorder, sparseness, irregularity, noise, and background of point clouds cause significant challenges in point cloud classification tasks. In such tasks, deep learning methods based on raw point cloud data have recently achieved good performance on simulated data. However, many methods experience problems when applied to realistic data containing much noise and complex background information. This paper proposes an end-to-end dual-input network (DINet) point cloud classification model based on deep learning. In the proposed model, a feature extractor obtains high-dimensional features, a feature comparator aggregates and disperses homogeneous and heterogeneous point clouds, respectively, in the feature space, and a feature analyzer completes the task. The two-channel data input facilitates a universal DINet framework that is flexible and extends to other models generalized across different datasets. DINet brings improvements in performance, presenting an overall accuracy value of 81.3% and average accuracy value of 79.6% in experiments conducted on the real-world ScanObjectNN dataset. The code for the proposed point cloud classification model is available at https://github.com/zhairf/DINet .

Highlights

Three-dimensional (3D) vision has become an area of active research as it provides more abundant information than two-dimensional (2D) vision
This paper proposes a dual-input network (DINet) framework for point cloud classification tasks in and a regularization function suitable for the framework, which can reduce the interference of noise and background on classification tasks
In this paper, we proposed the DINet framework, an endto-end deep learning architecture designed to improve the representational capacity of point cloud classification networks

Summary

INTRODUCTION

Three-dimensional (3D) vision has become an area of active research as it provides more abundant information than two-dimensional (2D) vision. R. Zhai et al.: Point Cloud Classification Model Based on a Dual-Input Deep Network Framework. In the field of 2D image classification, deep learning has engendered remarkable results [12]. As pointed out by Uy et al [30], many of the methods proposed for point cloud classification are successful with synthetic data but experience problems when applied to realistic data containing much noise and complex background information. This paper proposes a dual-input network (DINet) framework (see Fig. 2) for point cloud classification tasks in and a regularization function suitable for the framework, which can reduce the interference of noise and background on classification tasks. We propose a dual-input network (DINet) framework for point cloud classification. Evaluation results for the real-world open source dataset ScanObjectNN verify the efficacy of DINet

RELATED WORK

METHODS

EXPERIMENTS AND RESULT

Findings

CONCLUSION