| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 102 === In recent years, with the advancement of technologies, Wireless Visual Sensor Networks (WVSNs) become a popular research. Mobile object tracking problems were widely studied in the WVSNs. However, rare works discourse the quality of images captured by visual sensors. In this paper, our goal is to track a mobile object and obtain images of the object from arbitrarily angle and time. Most importantly, quality of each image is guaranteed. To achieve this goal, some basic sensing models are different from other research. First, Field of View (FoV) of the sensor is no longer a sector with a predefined angle and radius. Without setting the sensing range, we limit the FoV according to the Peak Signal-to-Noise Ratio (PSNR). Second, we assume the feature of the object cannot be known in advance. We intend to cover the entire object all the time to make sure that every detail of the object is not missing. Finally, we model the tracking problem as a set cover problem which considers both of the spatial and time domains at the same time. We propose two heuristic sensor selection schemes, sensor-based scheme and area-based scheme. According to our simulation results, while sensor-based scheme can reduce number of redundant nodes to be selected, area-based scheme can prolong more network lifetime. In conclusion, this work tackles the moving object tracking problem by tacking the image quality and full angle view into consideration and meanwhile achieve energy conservation.
|
|---|
