In this dissertation two new approaches have been introduced for the automatic detection of moving objects (such as people and vehicles) in video surveillance sequences. The first technique analyses the original video and exploits spatial and temporal information to find those pixels in the images that correspond to moving objects. The second technique analyses video sequences that have been encoded according to a recent video coding standard (H.264/AVC). As such, only the compressed features are analyzed to find moving objects. The latter technique results in a very fast and accurate detection (up to 20 times faster than the related work). Lastly, we investigated how different XML-based metadata standards can be used to represent information about these moving objects. We proposed the usage of Semantic Web Technologies to combine information described according to different metadata standards.