Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EDENet: Echo Direction Encoding Network for Place Recognition Based on Ground Penetrating Radar

Pengyu Zhang, Xieyuanli Chen, Yuwei Chen, Beizhen Bi, Zhuo Xu, Tian Jin, Xiaotao Huang, Liang Shen

TL;DR

This work tackles large-scale underground place recognition with Ground Penetrating Radar (GPR), where sparse features and dielectric variability challenge existing methods. It introduces EDENet, a directional echo encoding network that uses Learnable Gabor Filters (LGF) and a Direction Aware Attention (DAA) mechanism within Echo Direction Encoding (EDE) Blocks, augmented by a Shift Invariant Unit and multi-scale aggregation to produce compact global descriptors. The network is trained with a max-margin triplet loss to enable robust retrieval across changing underground conditions. On public GROUNDED and CMU-GPR datasets, EDENet significantly outperforms state-of-the-art methods while offering smaller model size and faster runtime, enabling real-time GPR-based PR in challenging environments.

Abstract

Ground penetrating radar (GPR) based localization has gained significant recognition in robotics due to its ability to detect stable subsurface features, offering advantages in environments where traditional sensors like cameras and LiDAR may struggle. However, existing methods are primarily focused on small-scale place recognition (PR), leaving the challenges of PR in large-scale maps unaddressed. These challenges include the inherent sparsity of underground features and the variability in underground dielectric constants, which complicate robust localization. In this work, we investigate the geometric relationship between GPR echo sequences and underground scenes, leveraging the robustness of directional features to inform our network design. We introduce learnable Gabor filters for the precise extraction of directional responses, coupled with a direction-aware attention mechanism for effective geometric encoding. To further enhance performance, we incorporate a shift-invariant unit and a multi-scale aggregation strategy to better accommodate variations in di-electric constants. Experiments conducted on public datasets demonstrate that our proposed EDENet not only surpasses existing solutions in terms of PR performance but also offers advantages in model size and computational efficiency.

EDENet: Echo Direction Encoding Network for Place Recognition Based on Ground Penetrating Radar

TL;DR

This work tackles large-scale underground place recognition with Ground Penetrating Radar (GPR), where sparse features and dielectric variability challenge existing methods. It introduces EDENet, a directional echo encoding network that uses Learnable Gabor Filters (LGF) and a Direction Aware Attention (DAA) mechanism within Echo Direction Encoding (EDE) Blocks, augmented by a Shift Invariant Unit and multi-scale aggregation to produce compact global descriptors. The network is trained with a max-margin triplet loss to enable robust retrieval across changing underground conditions. On public GROUNDED and CMU-GPR datasets, EDENet significantly outperforms state-of-the-art methods while offering smaller model size and faster runtime, enabling real-time GPR-based PR in challenging environments.

Abstract

Ground penetrating radar (GPR) based localization has gained significant recognition in robotics due to its ability to detect stable subsurface features, offering advantages in environments where traditional sensors like cameras and LiDAR may struggle. However, existing methods are primarily focused on small-scale place recognition (PR), leaving the challenges of PR in large-scale maps unaddressed. These challenges include the inherent sparsity of underground features and the variability in underground dielectric constants, which complicate robust localization. In this work, we investigate the geometric relationship between GPR echo sequences and underground scenes, leveraging the robustness of directional features to inform our network design. We introduce learnable Gabor filters for the precise extraction of directional responses, coupled with a direction-aware attention mechanism for effective geometric encoding. To further enhance performance, we incorporate a shift-invariant unit and a multi-scale aggregation strategy to better accommodate variations in di-electric constants. Experiments conducted on public datasets demonstrate that our proposed EDENet not only surpasses existing solutions in terms of PR performance but also offers advantages in model size and computational efficiency.

Paper Structure

This paper contains 25 sections, 9 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Underground Scene Representing
  4. Place Recognition (PR)
  5. Motivation
  6. Methodology
  7. Problem Definition
  8. Architecture Overview
  9. Echo Direction Encoding (EDE) Block
  10. Learnable Gabor Filters (LGF) Layer
  11. Direction Aware Attention (DAA)
  12. Shift Invariant Unit
  13. Feature Aggregation Layer
  14. Loss Function
  15. Experiment
  16. ...and 10 more sections

Figures (6)

  • Figure 1: GPR echoes with different dielectric constants.
  • Figure 2: Curve features obscured by interference.
  • Figure 3: The proposed framework consists of two modules. In EDENet, the sequential GPR data is compressed to generate a compact descriptor. In the position estimator module, we use a vector retrieval scheme to find the best matching.
  • Figure 4: The learnable Gabor filter.
  • Figure 5: Qualitative results of GROUNDED (Sunny/Rainy). Green and red boxes in indicate correct and wrong match.
  • ...and 1 more figures