Binary Graph Descriptor for Robust Relocalization on Heterogeneous Data

In this letter, we propose a novelbinary graph descriptor to improve loop detection for visual SLAM systems. Our contribution is twofold: i) embedding technique generating binary descriptors which conserve both spatial and histogram information extracted from images; ii) generic mean of combining multiple layers heterogeneous data into the proposed descriptor, coupled with matching geometric checking method. We also introduce an implementation our incremental Bag-of-Words (iBoW) structure that improves efficiency scalability, method interpret Deep Neural Network (DNN) results. evaluate system on synthetic real datasets across different lighting seasonal conditions. The outperforms state-of-the-art frameworks in terms relocalization precision computational performance, as well displays high robustness against cross-condition datasets.