Cost efficient deployment networked camera sensors

Using networked camera sensor nodes embedded with processing unit and transmission module provides new opportunities for many applications such as environmental monitoring and surveillance. Generally these applications have specific coverage requirements, and the cost of deployment is an important issue. In this dissertation we focus on the scenario that the camera sensors are randomly deployed, and propose various strategies to use them with minimal cost while achieving the desired coverage requirement of the applications.;We divided the problem into pre-deployment and post-deployment optimization problems. For the pre-deployment problem we focus on minimizing total cost while achieving the desired coverage. We develop analytical methods to derive the expected coverage of a single sensor as well as the joint coverage for a given number of homogenous and heterogeneous camera sensors. We also consider the case where obstacles may be present in the field. Following this we propose an adaptive sensor deployment strategy based on our analytical method and carry out simulation studies to validate the analytical results and to demonstrate that our strategy leads to near optimal overall cost.;For post-deployment problem we focus on optimizing power consumption to reduce unit time cost. We first propose an obstacle estimation scheme to support geometric based power optimization schemes. The estimation is done by detecting moving targets in the field. First an empirical detection model is built to define the visibility of a target to specific sensors, and then a histogram based mapping mechanism is proposed to assign confidence level for the presence of an obstacle for each cell in the field. Both simulation and hardware experiments are performed to validate that our approach works whether the target information is known or not.;We also propose a measurement based approach to identify and remove redundant camera sensors. We use the observation that two sensors have overlapped coverage area if they simultaneously detect a target to estimate whether a sensor is redundant. Based on these estimations, centralized and distributed algorithms are proposed to set the maximal number of sensors to idle state. The simulation results show that this approach leads to significant power saving.