Abandoned object detection with logical reasoning

Abandoned object detection is an essential requirement in many video surveillance contexts. We introduce an abandoned object detection tool based on a set of possible events and on a set of rules to act upon those events. This implementation is simple and reusable unlike existing techniques. It is implemented using a simple logical reasoning upon textual data, in contrast to image centric processing. Objects foreign to a usual environment are extracted using background subtraction. Results of blob detection and tagging process are passed to an abandoned object detector in a textual format. The abandoned object detector, which is an acyclic graph of asynchronously interconnected lightweight processing modules, evaluates the variations of speeds and inter-blob distances. By configuring several parameters according to the context, it generates an alert upon encountering such a scenario. We provide results of this implementation by applying it on PETS 2006 dataset.