LOW FALSE ALARM RATE VIDEO SECURITY SYSTEM USING OBJECT CLASSIFICATION

摘要

A video detection system (10) and method detects an intruder from video images of a scene. The method employs a recognition process to differentiate between humans and animals. The method is used only after possible false alarms resulting from identified effects of noise, aliasing, non-intruder motion occuring within the scene, and effects of global or local lighting changes. The object recognition process includes determining the regions containing a potential intruder, outlining and growing those regions to encompass all of potential intruders, determining a set of shape features from the region and eliminating possible shadow effects, normalizing the set, and comparing the normalized set with sets of features of humans and animals. This comparison produces a confidence level indicating a human intruder. An alarm is given for a sufficiently high confidence level. The possibility of a false alarm due to an animal or a non-identifiable object is also substantially eliminated.