Method for detecting intrusion on basis of video monitoring

Abstract

The invention discloses a method for detecting intrusion on the basis of video monitoring. The method includes acquiring front-end video stream data packets by the aid of FFMPEG and parallelizing acquired front-end video stream data by the aid of special CUDA (compute unified device architecture) decoding data interfaces; acquiring pedestrian classifiers by the aid of machine learning processes and establishing reference backgrounds by the aid of mathematical statistic processes. The method has the advantages that real-time online learning can be carried out, and the scene adaptability of algorithms can be improved; multiple targets can be tracked in real time by the aid of Kalman filter algorithms; pedestrian targets are tracked, the mass centers of the pedestrians are computed, target boundary violation warning effects can be realized if the mass centers are positioned in scribed regions, and video display can be carried out; the boundary violation targets can be quickly and accurately warned on the basis of automatic and artificial intelligent technologies, accordingly, intrusion of foreign targets can be effectively prevented, and the labor cost and the device-related prevention cost in perimeter prevention can be reduced to a great extent; the shortcoming of influence of change of light rays, swinging of trees and interference of insects can be overcome.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to an intrusion detection method based on video surveillance. Background technique [0002] With the development of society and the improvement of social public security needs, various security measures have emerged as the times require. The security issues of confidential key departments, perimeters, confidential rooms and other locations are mentioned by the security department as the top priority. The security of the security system is determined by the success of intrusion prevention in the perimeter, confidential rooms, and key departments. [0003] At present, the intrusion prevention methods commonly used in perimeters, confidential rooms, and key departments include: laser beams, infrared beams, electronic fences, vibrating optical cables, leaking cables, etc.; according to different scenarios, choosing different technical means can greatly improve To reduce fa...

AI Technical Summary

Problems solved by technology

[0008] Therefore, in the field of intrusion prevention, the methods mentioned above play a perceptual role, and usually need to be reviewed by connecting to video surveillance in the background; Personnel use machine vision methods to directly perceive on video, thereby reducing costs. However, the false alarm rate of intrusion detection using traditional machine vision methods is extremely high. Changes in light, swinging trees, and interference from insects all restrict The development and application of such methods

Images