Scene Context Based Semantic Segmentation for 3D LiDAR Data in Dynamic Scene

TL;DR

This paper introduces a graph neural network approach that leverages scene context for improved semantic segmentation of 3D LiDAR data in dynamic environments, outperforming previous methods.

Contribution

The proposed method uniquely encodes scene context using GNNs with automatic edge weight learning, enhancing segmentation accuracy in dynamic scenes.

Findings

8% improvement over unary CNN

17% improvement over normal GNN

Effective in both sparse and dense point clouds

Abstract

We propose a graph neural network(GNN) based method to incorporate scene context for the semantic segmentation of 3D LiDAR data. The problem is defined as building a graph to represent the topology of a center segment with its neighborhoods, then inferring the segment label. The node of graph is generated from the segment on range image, which is suitable for both sparse and dense point cloud. Edge weights that evaluate the correlations of center node and its neighborhoods are automatically encoded by a neural network, therefore the number of neighbor nodes is no longer a sensitive parameter. A system consists of segment generation, graph building, edge weight estimation, node updating, and node prediction is designed. Quantitative evaluation on a dataset of dynamic scene shows that our method has better performance than unary CNN with 8% improvement, as well as normal GNN with 17% improvement.