In this paper, we propose novel method for abnormal behavior detection in the video surveillance. The method constructs a normal behavior model using training samples, which makes possible application in a wide range of conditions and scenes. The method was tested in controlled and real conditions. The result shows that the method can be used to detect abnormal behavior in simple and crowded scenes.

We consider an algorithm of nonlinear spectral estimation based on the maximum-entropy method using the procedure of explicit obtaining of the Lagrange multipliers with reference to the problem of determining the time and frequency shifts of radio signals of communication systems. Modification of a method for constructing the uncertainty function on the basis of a nonlinear algorithm of spectral estimation with a fixed body of computations is proposed.

No abstract

No abstract