Visual Place Recognition of Robots via Global Features of Scan-Context Descriptors with Dictionary-Based Coding

Self-localization is a crucial requirement for visual robot place recognition. Particularly, the 3D point cloud obtained from laser rangefinders (LRF) applied to it. The critical part efficiency and accuracy of recognition robots based on cloud. current solution converting clouds 2D images, then processing these with convolutional neural network (CNN) classification. Although popular scan-context descriptor data can retain parts characteristics, its slightly low. This because image under adjacent label inclines be confusing. study reclassifies according CNN global features through feature extraction. In addition, dictionary-based coding leveraged construct retrieval dataset. experiment was conducted North-Campus-Long-Term (NCLT) dataset four-seasons conditions. results show that proposed method superior compared other methods without real-time Global Positioning System (GPS) information.