http://www.michaelbosse.com/papers/bosse_zlot_rss08_workshop.pdf

Keypoint design and evaluation for place recognition in 2D lidar maps

Approximate similarity search: A multi-faceted problem

Recognizing previously observed locations is a significant challenge in the areas of autonomous mapping and navigation. The place recognition problem arises in applications ranging from pose initialization and localization using prior maps, closing loops and relocalizing lost robots in simultaneous localization and mapping (SLAM), and merging maps collected at different times or by multiple robots. In this paper, we present and evaluate a general approach for efficiently recognizing and identifying previously observed areas. While we focus on two-dimensional laser point cloud maps, the methodology described is applicable to other sensing modalities and map representations. The maps we consider are constructed from large datasets taken over tens to hundreds of kilometers under regular driving conditions and road speeds in urban and suburban environments. At this scale, efficiency—especially in the case of online SLAM, which must be solved in real-time—is critical.

Place Recognition Using Keypoint Similarities in 2D Lidar Maps

Aggregate Nearest Neighbor (ANN) queries developing from Nearest Neighbor (NN) queries are the relatively new query type in spatial database and data mining An ANN queries return the object that minimizes an aggregate distance function with respect to a set of query points Because of the multiple query points, ANN queries are much more complex than NN queries For optimizing the query processing and improving the query efficiency, many ANN queries algorithms utilizes pruning strategies In this paper, we propose two points projecting based ANN queries algorithms which can efficiently prune the data points without indexing We project the query points into special "line", on which we analyses their distributing, then pruning the search space Unlike many other algorithms based on the data index mechanisms, our algorithms avoid the curse of dimensionality and are effective and efficient in both high dimensional space and metric space We conduct experimental studies using both real dataset and synthetic datasets to compare and evaluate their efficiencies

Efficient methods in finding aggregate nearest neighbor by projection-based filtering

Similarity search in high-dimensional metric spaces is a key operation in many applications, such as multimedia databases, image retrieval, object recognition, and others. The high dimensionality of the data requires special index structures to facilitate the search. A problem regarding the creation of suitable index structures for high-dimensional data is the relationship between the geometry of the data and the organization of an index structure. In this paper, we study the performance of a new index structure, called Divisive-Agglomerative Hierarchical Clustering tree (DAHC-tree), which reduces the effects imposed by the above liability. DAHC-tree is constructed by dividing and grouping the data set into compact clusters. We perform a rigorous experimental design and analyze the trade-offs involved in building such an index structure. Additionally, we present extensive experiments comparing our method against state-of-the-art of exact and approximate solutions. The conducted analysis and the reported comparative test results demonstrate that our technique significantly improves the performance of similarity queries.

DAHC-tree: An Effective Index for Approximate Search in High-Dimensional Metric Spaces

We explore using index structures for effective approximate and on-line nearest neighbor queries. While many index structures have showed to suffer from the dimensionality curse, we believe that indexes can still be useful in providing quick approximate solutions to the nearest neighbor queries. Moreover, the information provided by the indexes can provide certain bounds that can be invaluable for on-line nearest neighbor queries. This paper explores the idea of applying current R-tree based indexes to approximate and on-line nearest neighbors with bounds. We experiment with various heuristics and compare the trade-off between accuracy and efficiency. Our results are compared to locality sensitive hashing (LSH) and they show the effectiveness of the proposed scheme. We also provide guidelines on how this can be useful in a practical sense.

K. Kommeneni

Papers

Utilizing indexes for approximate and on-line nearest neighbor queries