Related Experiment Video
Updated: Aug 19, 2025
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
9.1K
STV-SC: Segmentation and Temporal Verification Enhanced Scan Context for Place Recognition in Unstructured
Xiaojie Tian1, Peng Yi1,2, Fu Zhang3
1Department of Control Science and Engineering, Tongji University, Shanghai 201804, China.
Sensors (Basel, Switzerland)
|November 26, 2022
Summary
This study introduces STV-SC, a new method for place recognition in simultaneous localization and mapping (SLAM) that uses segmentation and temporal verification. It improves mobile agent adaptability in unstructured environments by reducing false detections.
Area of Science:
- Robotics
- Computer Vision
- Artificial Intelligence
Background:
- Place recognition is crucial for Simultaneous Localization and Mapping (SLAM).
- LiDAR-based place recognition primarily uses geometric data, which is unreliable in unstructured environments.
- Existing methods struggle with environments dominated by unstructured objects.
Purpose of the Study:
- To propose STV-SC, a novel segmentation and temporal verification enhanced place recognition method.
- To improve the reliability and performance of place recognition in unstructured environments.
- To enhance the environmental adaptability of mobile agents.
Main Methods:
- Developed a range image-based 3D point segmentation algorithm.
- Implemented a three-stage loop detection process: two-stage candidate search and one-stage segmentation and temporal verification (STV).
- Utilized SLAM's time-continuous features within the STV process to identify and mitigate occasional mismatches.
Main Results:
- Demonstrated that the STV process effectively extracts structured objects and avoids outliers caused by unstructured environments.
- STV-SC achieves improved performance in unstructured environments, outperforming Scan context by 1.4-16% in recall rate at the same precision.
- The algorithm can operate online, enabling real-time applications.
Conclusions:
- STV-SC effectively addresses the limitations of geometric-based place recognition in unstructured environments.
- The proposed method enhances the robustness of SLAM systems by reducing false loop closures.
- STV-SC significantly improves the environmental adaptability and performance of mobile agents in challenging terrains.
