Robot/Cognitive System Research Department, Electronics & Telecommunication Research Institute, Yuseong-gu, Daejeon, Korea.
Sensors (Basel). 2012;12(4):3997-4015. doi: 10.3390/s120403997. Epub 2012 Mar 27.
In a wireless sensor network, sensors collect data about natural phenomena and transmit them to a server in real-time. Many studies have been conducted focusing on the processing of continuous queries in an approximate form. However, this approach is difficult to apply to environmental applications which require the correct data to be stored. In this paper, we propose a weather monitoring system for handling and storing the sensor data stream in real-time in order to support continuous spatial and/or temporal queries. In our system, we exploit two time-based insertion methods to store the sensor data stream and reduce the number of managed tuples, without losing any of the raw data which are useful for queries, by using the sensors' temporal attributes. In addition, we offer a method for reducing the cost of the join operations used in processing spatiotemporal queries by filtering out a list of irrelevant sensors from query range before making a join operation. In the results of the performance evaluation, the number of tuples obtained from the data stream is reduced by about 30% in comparison to a naïve approach, thereby decreasing the query execution time.
在无线传感器网络中，传感器实时收集有关自然现象的数据，并将其传输到服务器。许多研究都集中在以近似形式处理连续查询上。然而，这种方法很难应用于需要存储正确数据的环境应用程序。在本文中，我们提出了一种天气监测系统，用于实时处理和存储传感器数据流，以支持连续的空间和/或时间查询。在我们的系统中，我们利用两种基于时间的插入方法来存储传感器数据流，并通过使用传感器的时间属性减少管理元组的数量，而不会丢失任何对查询有用的原始数据。此外，我们提供了一种方法，通过在进行连接操作之前从查询范围中过滤出一组不相关的传感器，来降低处理时空查询时连接操作的成本。在性能评估的结果中，与简单方法相比，从数据流中获得的元组数减少了约 30%，从而减少了查询执行时间。
