Indexing of Hierarchically Organized Spatial-Temporal Data Using Dynamic Regular Octrees

Abstract

The paper is devoted to theoretical and experimental study of indexing methods as applied to spatial-temporal datasets appearing in different science and industry domains. For this purpose a general spatial-temporal data model is presented as a scene that admits hierarchically organized, heterogeneous spatial objects with individual temporal behaviors. For the model presented we argue the relevance of dynamic event-driven regular octrees as an underlying spatial-temporal indexing structure to a wide class of applications and prove its effectiveness for queries such as scene reconstruction, region search, and collision detection.

For hierarchically organized scenes a complementary generalization of the octrees is proposed. Its performance and memory consumption advantages over traditional structures are confirmed by carrying out a series of computational experiments with industry meaningful datasets originated from the construction modeling applications. Results of computational experiments substantiate theoretical conclusions and demonstrate possibilities of creating efficient applications under the conditions of permanently growing scales and complexity of spatial-temporal data.