ABSTRACT
The detection and segmentation of objects of interest, such as humans, vehicles, etc., is the first stage in automated visual event detection applications. It is always desirable to achieve a very high accuracy in the detection of targets with the lowest possible false alarm rates. We can characterize the domains of videos into four classes, depending on camera motion (static vs mobile) and the processing requirement for decision-making (offline vs online). Figure 4.1 shows these four categories with an example of each. Most surveillance videos are captured using static cameras. Whether processing is needed online or offline for this type of videos, traditional background subtraction techniques provide effective solutions for the detection and segmentation of targets in these scenarios.
