Detecting Moving Objects in Noisy Radar Data Using a Relational Database

In moving object databases, many authors assume that number and position of objects to be processed are always known in advance. Detecting an unknown moving object and pursuing its movement, however, is usually left to tracking algorithms outside the database in which the sensor data needed is actually stored. In this paper we present a solution to the problem of efficiently detecting targets over sensor data from a radar system based on database techniques. To this end, we implemented the recently developed probabilistic multiple hypothesis tracking approach using materialized SQL views and techniques for their incremental maintenance. We present empirical measurements showing that incremental evaluation techniques are indeed well-suited for efficiently detecting and tracking moving objects from a high-frequency stream of sensor data in this particular context. Additionally, we show how to efficiently simulate the aggregate function product which is fundamental for combining independent probabilistic values but unsupported by the SQL standard, yet.