Related Experiment Video
Updated: Jul 2, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
537
SC_LPR: Semantically Consistent LiDAR Place Recognition Based on Chained Cascade Network in Long-Term Dynamic
Summary
This study introduces a new method for robust place recognition (PR) using LiDAR data, effectively removing dynamic objects. The approach significantly enhances PR performance in challenging, ever-changing environments.
Area of Science:
- Robotics and Autonomous Systems
- Computer Vision
- Artificial Intelligence
Background:
- Dynamic objects in large-scale environments disrupt place recognition (PR) for mobile robots and autonomous vehicles.
- High-frequency changes caused by dynamic objects degrade scene appearance consistency over time.
- Robust PR is crucial for navigation and localization in real-world applications.
Purpose of the Study:
- To develop a novel, semantically consistent LiDAR place recognition (PR) method robust to dynamic objects.
- To improve the reliability of place recognition in large-scale, long-term dynamic environments.
- To introduce a new dataset for validating LiDAR semantic inpainting and PR methods.
Main Methods:
- A chained cascade network (SC_LPR) comprising a LiDAR semantic image inpainting network (LSI-Net) and a semantic pyramid Transformer-based PR network (SPT-Net).
- LSI-Net utilizes a coarse-to-fine generative adversarial network (GAN) with a gated convolutional autoencoder, incorporating Transformer blocks with mask attention and gated trident blocks.
- SPT-Net employs a pyramid Transformer encoder for global context encoding and an augmented NetVALD layer for feature aggregation.
Main Results:
- The proposed SC_LPR method achieved approximately 6% improvement in semantic inpainting performance.
- Place recognition performance in dynamic environments was enhanced by approximately 8% compared to baseline methods.
- A new LiDAR semantic inpainting dataset (LSI-Dataset) was created for method validation.
Conclusions:
- The developed SC_LPR method effectively eliminates the influence of dynamic objects for robust LiDAR place recognition.
- The integration of semantic inpainting and Transformer-based recognition offers a promising direction for PR in dynamic scenes.
- The LSI-Dataset and SC_LPR method provide valuable contributions to the field of autonomous navigation and robotics.
