Method for Frame Removal Detection in Static Camera Surveillance Video

Abstract

A method is proposed for passive protection of a surveillance camera video from a video fragment deletion attack. The method is based on the construction of local features for the samples of two consecutive frames, followed by a multilayer neural network classification. Post-processing and calculation of statistics based on the results of the classification make it possible to decide whether a given pair of frames is sequential or a number of frames were cut between them. Experiments show the efficiency of detecting the fact of a fragment removal even from stationary scenes, when such a deletion is visually imperceptible.