Querying Video Sensor Networks

A wireless sensor network is a resource constrained network with scarcity of computational capacity, memory, bandwidth and energy. The problem is further accentuated in multimedia wireless sensor networks. Raw video cannot be transferred over a wireless channel. Video compression is computationally intensive and even compressed video requires large amount of bandwidth for transfer. This entails the transfer of only specific targeted information over a video sensor network. This, in turn, requires specific queries for such networks along with low complexity techniques that generate query results requiring minimal data. The present study identifies the specific information needed by a video sensor network and proposes a video query language (VQL) for video sensor networks. VQL imposes a hierarchical structure on the sensor networks. The nodes responsible for tasks like compression, motion detection, data fusion and aggregation and routing are then specified. Relational algebra for VQL is proposed and discussed. The second part of the work focuses on techniques for motion detection with low complexity and minimal data. Simulation experiment is performed on video sequences for motion detection and image fusion in temporal dimension and results indicate that low amount of data is generated in detection of motion.