Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SIESEF-FusionNet: Spatial Inter-correlation Enhancement and Spatially-Embedded Feature Fusion Network for LiDAR Point Cloud Semantic Segmentation

Jiale Chen, Fei Xia, Jianliang Mao, Haoping Wang, Chuanlin Zhang

TL;DR

A novel spatial inter-correlation enhancement and spatially-embedded feature fusion network (SIESEF-FusionNet) is proposed in this paper, enhancing spatial inter-correlation by combining inverse distance weighting and angular compensation to extract more beneficial spatial information without causing redundancy.

Abstract

The ambiguity at the boundaries of different semantic classes in point cloud semantic segmentation often leads to incorrect decisions in intelligent perception systems, such as autonomous driving. Hence, accurate delineation of the boundaries is crucial for improving safety in autonomous driving. A novel spatial inter-correlation enhancement and spatially-embedded feature fusion network (SIESEF-FusionNet) is proposed in this paper, enhancing spatial inter-correlation by combining inverse distance weighting and angular compensation to extract more beneficial spatial information without causing redundancy. Meanwhile, a new spatial adaptive pooling module is also designed, embedding enhanced spatial information into semantic features for strengthening the context-awareness of semantic features. Experimental results demonstrate that 83.7% mIoU and 97.8% OA are achieved by SIESEF-FusionNet on the Toronto3D dataset, with performance superior to other baseline methods. A value of 61.1% mIoU is reached on the semanticKITTI dataset, where a marked improvement in segmentation performance is observed. In addition, the effectiveness and plug-and-play capability of the proposed modules are further verified through ablation studies.

SIESEF-FusionNet: Spatial Inter-correlation Enhancement and Spatially-Embedded Feature Fusion Network for LiDAR Point Cloud Semantic Segmentation

TL;DR

A novel spatial inter-correlation enhancement and spatially-embedded feature fusion network (SIESEF-FusionNet) is proposed in this paper, enhancing spatial inter-correlation by combining inverse distance weighting and angular compensation to extract more beneficial spatial information without causing redundancy.

Abstract

The ambiguity at the boundaries of different semantic classes in point cloud semantic segmentation often leads to incorrect decisions in intelligent perception systems, such as autonomous driving. Hence, accurate delineation of the boundaries is crucial for improving safety in autonomous driving. A novel spatial inter-correlation enhancement and spatially-embedded feature fusion network (SIESEF-FusionNet) is proposed in this paper, enhancing spatial inter-correlation by combining inverse distance weighting and angular compensation to extract more beneficial spatial information without causing redundancy. Meanwhile, a new spatial adaptive pooling module is also designed, embedding enhanced spatial information into semantic features for strengthening the context-awareness of semantic features. Experimental results demonstrate that 83.7% mIoU and 97.8% OA are achieved by SIESEF-FusionNet on the Toronto3D dataset, with performance superior to other baseline methods. A value of 61.1% mIoU is reached on the semanticKITTI dataset, where a marked improvement in segmentation performance is observed. In addition, the effectiveness and plug-and-play capability of the proposed modules are further verified through ablation studies.

Paper Structure

This paper contains 16 sections, 14 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep Learning-based Methods
  4. Encoding of Local Context Features
  5. Aggregation of Local Context Features
  6. Method
  7. Network Architecture
  8. Enhanced Local Spatial Encoding Module
  9. Spatially-Embedded Adaptive Pooling Module
  10. Experimental Verification
  11. Dataset Introduction
  12. Implementation Details
  13. Evaluation on the Toronto3D Dataset
  14. Evaluation on the SemanticKITTI Dataset
  15. Ablation Studies
  16. ...and 1 more sections

Figures (4)

  • Figure 1: The proposed SIESEF-FusionNet architecture.
  • Figure 2: The proposed reverse feature aggregation residual module is illustrated with the top part showing connections between these modules and the bottom part detailing the ELSE and SEAP modules.
  • Figure 3: Comparison of semantic segmentation results on the Toronto3D dataset.
  • Figure 4: Comparison of semantic segmentation results on the SemanticKITTI dataset. (a) Visualization of segmentation results for persons, trucks, and vegetation. (b) Visualization of segmentation results for bicyclists and cars, persons and roads.