Local-Global Feature Complementation Network for Visual Place Recognition

Visual place recognition (VPR) is crucial for robotic localization and navigation. Its key challenge lies in constructing feature representations that are robust to environmental changes. Existing methods typically adopt convolutional neural networks (CNNs) or vision transformers (ViTs) as feature extractors. Notably, these architectures excel in different aspects—CNNs are effective at capturing local details, while ViTs are better suited for modeling the global context. Thus, leveraging the strengths of both networks simultaneously is difficult. To address this issue, we propose the Local-Global Feature Complementation Network (LGCN) that integrates a parallel CNN-ViT hybrid architecture with the Dynamic Feature Fusion Module (DFM) for VPR. DFM performs dynamic feature fusion through the joint modeling of spatial and channel-wise dependencies. Furthermore, to enhance the expressiveness and adaptability of the ViT branch for VPR tasks, we introduce lightweight frequency-spatial fusion adapters into the frozen ViT backbone. These adapters enable task-specific adaptation using a controlled parameter overhead. Extensive experiments on multiple VPR benchmark datasets demonstrate that the proposed LGCN consistently outperforms existing approaches in terms of localization accuracy and robustness, validating its effectiveness and generalizability.