Treffer: Deep spatial-frequency fusion for loop closure descriptor construction.

Loop closure detection (LCD) is essential in simultaneous localization and mapping, as it mitigates accumulated errors and ensures consistent long-term mapping. However, existing methods often face limitations in adaptability to varying light detection and ranging (LiDAR) configurations and inefficient utilization of point cloud data. To address these challenges, we propose SF-LCD, a deep learning framework that constructs highly discriminative loop closure descriptors through spatial-frequency fusion of raw LiDAR point clouds. Our approach employs parallel convolutional neural networks to independently extract spatial geometric features from depth images and multi-scale frequency features via wavelet transform. A novel self-attention-based fusion module adaptively balances the spatial and frequency components through attention weights, thereby enhancing robustness to environmental variations. Experiments on public datasets demonstrate that SF-LCD achieves consistently strong performance across various LiDAR beam configurations. [ABSTRACT FROM AUTHOR]

Copyright of Measurement Science & Technology is the property of IOP Publishing and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)