Due to the advances in digital information technologies and dramatic drop in the prices of storage devices, the installation of visual surveillance systems (VSS) has become relatively inexpensive. However, maintaining the staff to monitor these CCTV sources is still a costly affair. Moreover, as the end users are becoming more aware of the vision based technologies, there is ever growing demand for advanced surveillance systems which can not only model and detect complex interesting events but can also provide intelligence to improve their operational management process. In this thesis we investigated the three main aspects of visual surveillance systems (event modelling, event detection and event mining) and propose the framework in which these different aspects of surveillance systems complement each other. The research contributions presented in the thesis mostly fall under the event mining and event modelling aspects of visual surveillance systems. Most of the previous work for mining multimedia events is based upon discovering/detecting already known abnormal events or deals with finding frequent event patterns. In contrast, in this thesis we present a framework to discover unknown anomalous events associated with a frequent sequence of events (AEASP); that is to discover events which are unlikely to follow a frequent sequence of events. This information can be very useful for discovering unknown abnormal events and can provide early actionable intelligence to redeploy resources to specific areas of view (such as PTZ camera or attention of a CCTV user). Discovery of anomalous events against a sequential pattern can also provide business intelligence for store management in the retail sector. The proposed event mining framework also takes the temporal aspect of AEASP into consideration, that is to discover anomalous events which are true for a specific time interval only and might not be an AEASP over a whole time spectrum and vice versa. To confront the process/memory expensive process of searching all the instances of multiple sequential patterns in each data sequence a dynamic sequential pattern search mechanism (DSPS_SM) is also introduced. Different experiments are conducted to evaluate the proposed AEASP mining algorithm's accuracy and performance. Next, we proposed an event mining framework to automate the process of generating appearance models of real world entities by utilising the results of already detected events and text streams of multimedia events. A comprehensive problem definition and entity appearance model generation framework is presented. To validate the proposed entity appearance model generation concept, we implemented the proposed algorithm and conducted the experiments by using "Columbia University Image Library (COIL-I 00)" object database [I). To address the event modelling aspect of surveillance systems we extended and modified the event description framework (EDF) presented in [2] and proposed the extended version of it (EDFE). EDFE not only increased EDF's capability to model complex multimedia events but also facilitated the event detection and event mining processes. Modelled events (generated by EDFE) and the event detection process is evaluated using sequences generated in laboratory and also from realistic surveillance environment, the results of experiments are then analysed using precision and recall measures.