Human-inspired place recognition improves visual SLAM speed and accuracy
HuMemSLAM: Efficient Human-Inspired Semantic Place Recognition for Robust Visual SLAM
RoboticsComputer Vision and Pattern Recognition
Summary
Robots and autonomous systems must recognize places to build maps and locate themselves accurately. Traditional methods use simple visual clues but can get confused when different places look alike or the same place looks different over time. The authors created HuMem-VPR, a new approach inspired by how humans remember places, combining immediate visual input with contextual understanding. This method was faster and more accurate than other leading systems and improved an existing SLAM system’s ability to recognize places correctly.
What this means in practice
- •For robotics engineers: Enhance autonomous robots’ map-building and localization accuracy while reducing processing delay using a human-inspired place recognition system integrated into SLAM workflows.
- •For autonomous vehicle developers: Improve vehicle visual localization robustness in complex or changing environments by deploying a faster, context-aware place recognition method within onboard SLAM systems.
Authors
Mayowa Adebambo, Sebastian Donnelly, Armand Amaritei, Andrew Bradley, Alexander Rast
Abstract
Autonomous systems require reliable place recognition for efficient and effective simultaneous localisation and mapping (SLAM). Traditional geometric visual SLAM approaches rely on low-level features and geometric consistency, but remain vulnerable to perceptual aliasing, where different places appear similar, and perceptual variation, where the same place appears different. Although semantic SLAM and modern learned visual place recognition (VPR) methods improve robustness under challenging perceptual conditions, real-time deployment requires both high retrieval accuracy and low latency. Inspired by human memory and perception, we propose HuMem-VPR, which exploits the bidirectional relationship between bottom-up perceptual evidence and top-down contextual reasoning to achieve high-level place understanding. We further introduce HuMemSLAM, the integration of HuMem-VPR with ORB-SLAM3. HuMem VPR achieved the highest aggregate retrieval accuracy on the real-image benchmark, competitive accuracy on the CARLA benchmark, and approximately two to three times lower latency than the evaluated state-of-the-art VPR methods. Across the evaluated dataset families and online experiments, HuMemSLAM substantially improved integrated Recall @1 over ORB-SLAM3's native retrieval while reducing the proposals submitted to its geometric backend.