Detection of Moving Objects in Surveillance Video by Integrating Bottom-up Approach with Knowledge Base

In the modern age, where every prominent and populous area of a city is continuously monitored, a lot of data in the form of video has to be analyzed. There is a need for an algorithm that helps in the demarcation of the abnormal activities, for ensuring better security. To decrease perceptual overload in CCTV monitoring, automation of focusing the attention on significant events happening in overpopulated public scenes is also necessary. The major challenge lies in differentiating detecting of salient motion and background motion. This paper discusses a saliency detection method that aims to discover and localize the moving regions for indoor and outdoor surveillance videos. This method does not require any prior knowledge of a scene and this has been verified with snippets of surveillance footages.