Advanced indexing methods for large spatial data in complex dynamic scenes

• Main Authors: V. A. Zolotov, V. A. Semenov
• Format: Article
• Language: English
• Published: Ivannikov Institute for System Programming of the Russian Academy of Sciences 2018-10-01
• Series: Труды Института системного программирования РАН
• Subjects: spatial indexing, multidimensional data, dynamic scene modeling
• Online Access: https://ispranproceedings.elpub.ru/jour/article/view/964

This paper is dedicated to review of recent methods for indexing of multidimensional data and their use for modeling of large-scale dynamic scenes. The problem arises in many application domains such as computer graphics systems, virtual and augmented reality systems, CAD systems, robotics, geographical information systems, project management systems, etc. Two fundamental approaches to the indexing of multidimensional data, namely object aggregation and spatial decomposition, have been outlined. In the context of the former approach balanced search trees, also referred as object pyramids, have been discussed. In particular, generalizations of B-trees such as R-trees, R*-trees, R+-trees have been discussed in conformity to indexing of large data located in external memory and accessible by pages. The conducted analysis shows that object aggregation methods are well suited for static scenes but very limited for dynamic environments. The latter approach assumes the recursive decomposition of scene volume in accordance with the cutting rules. Space decomposition methods are octrees, A-trees, bintrees, K-D-trees, X-Y-trees, treemaps and puzzle-trees. They support more reasonable compromise between performance of spatial queries and expenses required to update indexing structures and to keep their in concordant state under permanent changes in the scenes. Compared to object pyramids, these methods look more promising for dramatically changed environments. It is concluded that regular dynamic octrees are most effective for the considered applications of modeling of large-scale dynamic scenes.