Energy efficient processing and distributed power management for wireless video sensor networks

The successful proliferation of multimedia-enabled devices and significant advances in Very Large Scale Integration (VLSI) technology has spawned new research efforts in migrating video processing onto ever smaller and cheaper devices. Additionally, as Wireless Sensor Networks (WSN) continue to improve in functionality with memory and processing resources becoming cheaper and more power efficient, future WSN systems will form Video Sensor Networks (VSN) containing video as a primary sensing mode. This dissertation focuses on both node-level and network-level issues to achieve power efficiency for future VSN systems. These new techniques include: (1) A Region of Interest (ROI) video processing technique is introduced which limits processing to a portion of a video frame in an attempt to successfully track a dominant moving object within a single camera's Field Of View (FOV) while saving on considerable processing requirements. (2) An Adaptive Data Partitioning (ADP) algorithm is described which splits each video frame into multiple regions and assigns each region to a different Processing Element (PE) to parallelize the processing task. (3) A set of Coordinated Distributed Power Management (CDPM) policies are proposed which extend well-known local power management policies to include information from multiple nodes within the VSN.; A detailed VSN simulator has been used in the experiments to help analyze performance/resource-savings tradeoffs and prototype VSN nodes have been built to physically verify performance of many of the techniques included herein.