| 摘要: | 近年來，由於科技及硬體技術的進步，使得體積小、記憶容量大、低成本、多功能、省電、且具無線與感測能力的感測器得以實現。這亦使無線感測網路的應用日趨成熟，並被廣泛應用在許多的領域，例如：災害回報、軍事偵測與監控、位置追蹤、環境監控等多種領域。基於上述的應用領域都有使用感測器來幫助偵測、追蹤某些特定事件，所以我們的研究議題是在無線感測網路中設計連續型物體的偵測與追蹤協定。 目前大部分物體的偵測與追蹤協定其設計重點是針對個體型物體的偵測與追蹤，例如：入侵者、坦克、交通工具等個體型物體。這些個體型物體同通常只會被其所在周圍數個感應器所感測。因此，個體型物體偵測與追蹤協定的設計重點在於，感測到物體的事件節點之間要如何協同計算出物體位置或如何有效率的將感測到物體的位置送到後端使用者。而連續型物體不同於個體型物體的地方在於：連續型物體通常會使得大量的感測器同時偵測的物體的出現，且物體的形狀、體積、位置會隨著時間改變，例如：森林大火、有毒氣體擴散，群居動物遷移、軍隊的行進等等。因此，連續型物體偵測與追蹤協定的設計所要考慮的重點在於如何利用最少的資訊來評估物體的位置以減少物體資訊的回報量，進而減少通訊所造成的電量耗損。 在連續型物體偵測協定設計上，我們提出了兩個協定：第一個協定為，以網路拓墣控制為基礎的分散式連續型物體偵測協定；第二個協定為，以涵蓋方式為基礎的分散式連續型物體偵測協連續型物體偵測協定。在第一個所提出的協定中，我們首先利用三角劃分來減少每一個感測器鄰近節點的數目，進而將事件節點中最靠近物體邊界的節點挑選成邊界節點。被挑選出的邊界節點將會回報它們自己的位置資訊給使用者。而使用者藉由所接收的資訊，可以評估出所有事件節點的位置。在這個協定中，我們同時提出兩個規則來進一步的減少邊界節點的數量，且所提出的規則將不會花費額外的通訊成本。 由於，邊界節點的數量主要是受到連續型物體的邊長所影響。也就是說，無論連續型物體的形狀或體積如何，邊長較長的連續型物體將會有較多的邊界節點。因此，我們提出第二個以涵蓋方式為基礎的分散式連續型物體偵測協連續型物體偵測協定，來改善這個問題。在這個協定中，我們找出所有事件節點中的核心節點。核心節點的位置與跳數(hop-count)半徑資訊可以表達一個涵蓋區域。而所有核心節點的涵蓋區域可以涵蓋出所有事件節點的位置。因此，當使用者收到所有核心節點的回報資訊，將可以評估出連續型物體的位置。由於，核心節點的數量少於邊界節點，所以藉由核心節點來回報資訊給使用者將可以有效率的減少回報物體資訊的通訊量。 在無線感測網路中，感測節點之間需要利用節點編號(node ID)來進行通訊。且在我們設計的連續型物體的偵測與追蹤協定中，需要每個感測節點都要有一個獨一無二的節點編號。因此，在本論文中我們同時設計了一個節點編號分配協定。藉由我們所設計的協定，在無線感測網路佈置後每一個感測節點將可以快速的取得一個獨一無二的節點編號。透過實驗的測試分析我們所設計協定的優越性都得到驗證。 Recently, scientific progress and hardware technology have made possible the development of wireless sensor nodes in small sizes with large memory capacity, low cost, and the multi-function, which allow a wide range of applications for wireless sensor network, such as in battlefield surveillance, calamity reporting, biomedicine, hazardous environment exploration, environmental monitoring, and position tracking. Since these applications must use sensor nodes to detect and track some specific events, the object detection and tracking protocol is a basic function for these applications. In this dissertation, our research focuses on detection and tracking of continuous objects in wireless sensor networks. Most of the previous researches on object detection and tracking focus on individual objects, such as intruder, tank, vehicle, etc. The individual object can usually be detected by several neighboring sensor nodes, called event nodes. Therefore, the collaboration of event nodes and efficient transmission of object information to sink nodes are considered in designing individual object detection and tracking protocols. Compared with the individual object, the continuous object is usually larger in size and can diffuse, increase in size, or split into multiple continuous objects, such as forest fire, noxious gas, the social animal moving and the army advancing, which imply that there should be a lot more event nodes. Hence, the main challenge and goal for designing the continuous object detection and tracking protocol is to reduce the number of reporting information to the sink node for energy conservation in communication. Therefore, individual object detection and tracking protocols are unsuitable for continuous object detection and tracking. In this dissertation, we propose two continuous object detection and tracking protocols: a scalable topology-control-based continuous object detection and tracking (STCB) protocol and a coverage-based continuous object detection and tracking (CBDT) protocol. In the STCB protocol, the event nodes close to the boundary of object will be selected as boundary nodes in a quick distribution manner. Moreover, the location of these selected boundary nodes can represent the location of all event nodes. Therefore, only boundary nodes report their location information to sink nodes; the sink nodes can estimate the location information of continuous objects. Two rules for reducing the boundary node, without spending extra communication cost, are also discussed in the STCB protocol. The number of selected reporters in a continuous object is proportional to the length of object boundary no matter what size or how sharp the object is. In CBDT protocol, one or several core nodes will be determined from the boundary nodes for reporting continuous object information. The location of the core nodes and their determined hop count can represent all the locations of the event nodes. Therefore, only the core nodes report their information to the sink node; the communication cost in continuous object detection and tracking can be further reduced. In wireless sensor network, each sensor node needs a unique node ID for communication. A sensor node with a unique node ID would be assumed in our object detection and tracking protocol. Therefore, we also proposed a new node ID assignment protocol for each sensor node that obtains a unique node ID quickly in the network initialization. Through the simulations, the proposed protocols are verified to be both efficient and cost-effective. |
