SpatioTemporalDBKit: Database Extensions for Moving-Object and Event-Based Analytics
Main article
Abstract
Spatial and temporal data streams generated by GPS-equipped vehicles, IoT sensor networks, urban mobility systems, and satellite tracking platforms collectively represent one of the fastest-growing categories of operational data in modern computing. Yet the relational database ecosystem still lacks a consolidated, schema-documented, and publicly available toolkit that transforms raw moving-object streams into analytically ready knowledge structures. This paper presents SpatioTemporalDBKit (STDBKit), a modular database extension library built on top of PostgreSQL and PostGIS that provides trajectory storage, spatiotemporal indexing, event detection, and window-based analytical functions for moving-object and event-based analytics. STDBKit introduces six normalized relational tables with complete field dictionaries and foreign-key integrity, a configurable five-stage ingestion pipeline covering coordinate-reference-system normalization, noise filtering, trajectory segmentation, Douglas-Peucker compression, and R*-tree or SP-GiST spatial indexing. An extension function layer delivers over 40 PL/pgSQL and Python user-defined functions for trajectory interpolation, spatial join, event-sequence detection, and sliding-window aggregation. Three benchmark experiments are reported: query latency scaling across dataset sizes from 0.5 M to 50 M GPS points; index size and build-time comparison against PostGIS and MobilityDB; and trajectory compression accuracy measured by Synchronized Euclidean Distance. On a 50 M-point benchmark corpus, STDBKit SP-GiST indexing reduces index size by 51% and build time by 47% relative to a vanilla PostGIS R-tree configuration, while achieving a median range-query latency of 38 ms. All source code, schema scripts, seed data, and Docker images are released under Apache 2.0.
