| Camera sensor networks have many applications in practice and have received much research attention in the past few years. In terms of coverage, they are dramatically different from traditional sensor networks. However, most existing works on sensor networks have been using relatively simple models to characterize coverage and few results can be used to solve the coverage problems in camera sensor networks. One unique property of camera sensors is that different cameras from different positions can form distinct views of the object, and a good coverage often demands views from all around the object. This distinction makes the coverage verification and deployment methods designed for traditional sensor networks unsuitable. Also, as the sensing range of a camera sensor is not disk like but directional, it may steer around to monitor different areas, which brings more flexibility in coverage but also more complex in designing a good scheduling protocol. Meanwhile, emerging applications such as mobile image sensing that is based on camera sensors need to be carefully studied.;The main purpose of this dissertation is to address new challenges in the coverage problems of camera sensor networks and provide deep insight into many other related issues. First, we define a new coverage model called full-view coverage, which precisely characterize the uniqueness of coverage in camera sensor networks. Under this framework, we design a series of efficient methods for coverage detection, camera sensor deployment (in both random and deterministic environment), and provide rigorous analysis on the accuracy and efficiency of the proposed methods.;Second, we extend our research into the scenario of barrier coverage. Two important types, i.e., weak barrier and strong barrier coverage have been redefined with the consideration of the aforementioned features of camera sensors. In both cases, the object traversing a monitored field are guaranteed to be detected by the camera sensor networks on its face. While strong barrier coverage allows more flexibility on how the object chooses its path and facing direction, weak barrier coverage demands less camera sensors. For both cases, novel algorithms are proposed to find and construct camera sensor barriers, and the performance is validated by rigorous theoretical analysis and simulations.;Third, we show how to schedule a camera sensor networks with steering capability to minimize service delay (gap). The problem studied is a generic case where directional sensors like cameras can steer around and periodically cover different targets at different time. By driving directional sensors rotate and cover multiple areas, the sensing range can be expanded and the total number of sensors required can be reduced. The proposed optimization problem is proved to be NP-hard and both centralized and distributed protocols are proposed with proved performance bound.;Finally, we expand our study into an application of camera sensors by considering smartphone image sensing. Photos obtained by smartphone users can assist in situations like disaster recovery. A critical challenge here is the mismatch between the huge amount of available photos at the user end and the limited communication and computation resource to transfer to and process the crowdsourced photos at the server. A key is to utilize the metadata associated with each photo including the smartphone camera's parameters and other geographical information, and model the photos' usefulness by utility, similar but more flexible than the original full-view coverage model. Optimization problems regarding tradeoffs between resources and photo coverage are introduced and efficient solutions are proposed. Performance of the solutions are rigorously proved. Furthermore, a testbed evaluation and extensive simulation demonstrate the effectiveness of the system. iv. |
