Kluster: Application of k-means clustering to multidimensional GEO-spatial data

Abstract

There have been several developments both in hardware and softwares for capturing and processing geographic information. The results yield by the processing of geo-spatial data helps for efficient planning and strategic decision making. Due to the amount of geo-spatial data captured and stored every day, it has become of utmost importance that the data be processed and the results be delivered in time. With the increasing amount of data, application of distributed and parallel frameworks which have been proven capable of processing large amounts of data in other domains has become necessary. These frameworks have to be extended to support geo-spatial data and related geo-processing operations. In this paper, we have extended the regular k-means clustering to support processing of multi spectral geo-spatial raster data over Hadoop. By performing clustering on multiple dimensions (multiple spectrums) simultaneously over a distributed processing framework, Apache Hadoop, our approach allows detailed clustering. Our approach in addition to considering n-dimensional data also allows processing it by utilizing heterogeneous and distributed resources for faster processing and delivery results.