Processing information in real-time that is coming from video streams is very challenging due to uncertainties such as clutter (correct identifying objects that are however of no interest to the observer) and noise (errors ad rather probabilistic disturbances). This thesis details an approach for automatic novelty detection, single and multiple object(s) identification in video streams. It is based on a method that provides recursive density-estimation. Instead of the usually used Gaussian, it is using a Cauchy type of kernel. It should be emphasized that the introduced approach is computationally efficient and can be implemented in real-time; in addition, it can be extended for multiple objects identification and tracking as detailed in this thesis. For trajectory analysis of moving objects and anomaly detection, multi-feature spaces are used to increase the accuracy of detecting deviant behaviors. Evolving clustering is also used for real-time trajectory clustering purposes. Self-evolving parameter-free rule-based controller is proposed in this thesis. The proposed controller can start with no pre-defined fuzzy rules or control variables. It learns from its own action during the control process. It does not use the explicit model or explicit membership functions. It combines the concept of parameter-free data density fuzzy rule-based systems with newer concepts of self-evolving controllers. It is possible to generate a parameter free control structure based on the data density and selecting representative focal points from the control surface. The illustrative results aim primarily proof of concept.