SpatialHadoop: For OpenStreetMap Data

With the change of time information related to geography and volunteered geography also changes. In this way extraction of spatial patterns from crowdsourced data has become most valuable for service suppliers. These patterns represent the spatial features of the co-related objects. The existing approaches used Dijkstras algorithm and Euclidean distance to find spatial patterns which can not compute accurately. Crowdsourced data is growing on daily basis through mobile phones, road networks and remote sensors. In order to process this type of large data set is also becoming difficult. In this research work we have proposed a system to process crowdsourced data taken from OpenStreetMap to mine the useful patterns using SpatialHadoop. SpatialHadoop has used Pigeon script, a spatial extension to Pig that is a high level language. These patterns will assist service providers to offer different sites based on facilities. In this extraction method, spatial data is loaded into the system and filtered for nodes, ways and relations. The filtered data is used for the mining process by using kNN joins. After this the evaluation of multiple resolution pruning filter with spatial datasets are generated using argument values. The different data sets have been checked using this methodology to extract the spatial patterns. This technique is compared with PostgreSQL and it is observed that SDM has provided more efficient results. The result obtained from experiment has shown the performance of our system that is better in comparison to the already existing systems in consideration of efficiency, speed and accuracy that rely on a network.