تشکیل تصویر پانوراما با استفاده از اطلاعات شبکه های حسگر دوربین

wireless sensor networks (wsns) are one of the most interesting consequences of innovations in different areas of technology including wireless and mobile communications, networking, and sensor design. these networks are considered as a class of wireless networks which are constructed by a set of sensors. a large number of applications have been proposed for wsns. besides having numerous applications, different types of issues must be addressed regarding the sensor networks such as placement, data routing, power consumption, and security. wireless camera sensor network (wcsn) is constructed by a set of small and low cost sensor nodes which can produce images or videos from the sensing area. the images are then processed and transmitted to other sensors or to a central base station. these networks are highly applicable in security surveillance, environment monitoring, industrial process control, and person locator services. due to their broad applications, these networks have been widely addressed in recent works. view enhancement and field of view expansion are some applications of camera sensor networks. these networks can also enable multi-resolution views. furthermore, the ability of object tracking in a dramatically large area is another advantage of these networks. the main difference between the wcsn and other types of wsn is the type of the sensors which are employed in wcsns. regular (non-camera) sensors collect scalar data from the environment, while camera sensors make images of the surrounding environment. camera sensor networks are categorized as homogeneous and heterogeneous ones. homogeneous wcsns are constructed by exactly similar sensor nodes, while in heterogeneous networks various types of sensors can work together. in heterogeneous networks not only different types of camera sensors may be employed but also, non-camera sensors may be used as well. an interesting and versatile application of wcsns is in security, surveillance and environment monitoring. therefore, seamless coverage of a vast area is an important issue in these applications. it has been shown that in regular sensor networks, coverage is an np-complete problem. hence, heuristics have been proposed for indoor environments such as shopping centers, hospitals, and airports. the goal of this research is to apply optimization, image processing, and machine vision techniques for better use of the network’s resources in order to generate an image of the intended scene. therefore, at each time step, a group of sensors should be selected to cover the intended area such that the life of the network could be maximized. also, since the sensors need to communicate with a central node, issues such as routing should be considered. using machine vision techniques a panoramic image of the scene under surveillance is generated. in this thesis four groups of algorithms are presented for better selection of sensors while considering the routing problems. these groups of algorithms are classified as priority, visual neighbouring, communication neighbouring, and evolutionary methods. simulation results reveal that priority based, visual neighbouring ones, and evolutionary methods are suitable when a central decision system exists. on the other hand, when the decision making is distributed among the sensors, algorithms based on communication neighbouring are more suitable. evolutionary methods show an overall better performance in regards to the use of network resources.