In robotics applications such as SLAM (Simultaneous Localization and Mapping), loop closure detection is an integral component required to build a consistent topological or metric map. This paper presents an appearance based loop closure detection mechanism titled `IBuILD' (Incremental bag of BInary words for Appearance based Loop closure Detection). The presented approach focuses on an online, incremental formulation of binary vocabulary generation for loop closure detection. The proposed approach does not require a prior vocabulary learning phase and relies purely on the appearance of the scene for loop closure detection without the need of odometry or GPS estimates. The vocabulary generation process is based on feature tracking between consecutive images to incorporate pose invariance. In addition, this process is coupled with a simple likelihood function to generate the most suitable loop closure candidate and a temporal consistency constraint to filter out inconsistent loop closures. Evaluation on different publicly available outdoor urban and indoor datasets shows that the presented approach is capable of generating higher recall at 100% precision in comparison to the state of the art.

This paper presents an extension of the standard occupancy grid for 3D environment mapping. The presented approach adds a fusion process after the occupancy update which modifies the resolution of the grid cells in an incremental manner. Consequently, the proposed approach requires fewer grid cells for 3D representation in comparison to a standard occupancy grid. The resolution adaptation process is based on the occupancy probabilities of the grid cells and leads to the relaxation of the cubic grid cell assumption common to most 3D occupancy grids. The aim of this paper is to show the advantage of the proposed incremental fusion process which leads to the approximation of the 3D environment using rectangular cuboids. Evaluation on a large scale dataset and comparison to the state of the art shows that the proposed approach has faster access time for all occupied grid cells and requires a smaller number of cells for 3D environment representation.

In this paper a novel approach for 3D environment mapping using registered robot poses is presented. The proposed algorithm focuses on improving the quality of robot generated 3D maps by incorporating the uncertainty of 3D points and propagating it into the normal vectors of surfaces. The uncertainty of normal vectors is an indicator of the quality of the detected surface. A controlled random search algorithm is applied to optimize a non-convex function of uncertain normal vectors and number of clusters in order to find the optimal threshold parameter for the segmentation process. This approach leads to an improved cluster coherence and thus better maps.

In this paper, an online spatial and appearance based loop closure algorithm is presented. The approach is based on a graph matching formulation using Position Invariant Robust Features (PIRF), extending previous approaches based on PIRF by incorporating spatial information. The vertices of the graph represent visual words/features and edges represent metric information with uncertainty since the spatial distances observed between visual words are prone to errors. This method is capable of detecting loop closure in urban environments based on visual appearance as well as spatial layout of matched visual features. A vocabulary is built in an online and incremental manner, also storing spatial distances between visual words. The algorithm is capable of assigning loop closure with higher confidence values and a higher recall rate while maintaining precision compared to approaches where only visual appearance methods are used. We evaluate this approach on a publicly available dataset and present experimental results.