| Summary: | 博士 === 國立臺灣大學 === 資訊工程學研究所 === 99 === Camera network have been widely used in visual surveillance applications, such as airport or railway security, traffic monitoring, and etc. The main benefit of multi-camera system is that it can monitor the activities of targets over a large area. However, to security guards or users, the difficulty of monitoring such a system increases with the increase of cameras, especially when the events happen among multiple cameras. In this dissertation, we investigate two major tasks of monitoring in the command center display. One is to track targets in a camera network with computer automation. The other is to develop displaying techniques to help users to monitor the events in a camera network more easily.
First, to track targets across networked cameras, we focus on the situations where the view fields of cameras are not necessarily overlapping each other. One of the major problems of tracking across non-overlapping cameras is to learn the spatio-temporal relationship and the appearance relationship, where the appearance relationship is usually modeled as a brightness transfer function. Traditional methods learning the relationships by using either hand-labeled correspondence or batch-learning procedure are applicable when the environment remains unchanged. However, in many situations such as lighting changes, the environment varies seriously and hence traditional methods fail to work. In this dissertation, we propose an unsupervised method which learns adaptively and can be applied to long-term monitoring.
Second, when monitoring the tracking activity in the camera network, the traditional surveillance systems usually switch the main camera view from one to another directly, but it makes users difficult to be aware of the trajectory of the target in the environment when switching views many times. In this dissertation, we propose a novel egocentric view transition approach, which synthesizes the virtual views during the period of switching cameras and eases the mental effort for users to understand the events. An important property of our system is that it can be applied to the situations of where the view fields of transition cameras are not close enough or even exclusive.
Finally, for large-scale and high-resolution monitoring, we proposed a multi-resolution display with steerable focus, e-Fovea,. Large-scale and high-resolution monitoring systems are ideal for many visual surveillance applications. However, existing approaches have insufficient resolution and low frame rate per second, or have high complexity and cost. We take inspiration from the human visual system and propose a multi-resolution design, e-Fovea, which provides peripheral vision with a steerable fovea that is in higher resolution. In this dissertation, we further present two user studies, with a total of 36 participants, to compare e-Fovea to two existing multi-resolution visual monitoring designs. The user study results show that for visual monitoring tasks, our e-Fovea design with steerable focus is significantly faster than existing approaches and preferred by users.
|
|---|
