A method and device for detecting camera interference

Abstract

The invention discloses a method and apparatus for detecting interference of a camera. The method comprises the following steps: step 1, calculating an image entropy of a current video frame, and initially judging whether the camera has interference through the size of the image entropy; step 2, if it is judged that the camera has no interference in step 1, obtaining a continuously changing area of the camera within the latest N frames, and judging the presence of interference of the camera through the size of the continuously changing area; step 3, creating or regularly updating a reference background; and step 4, when the reference background exists, comparing the difference between the current frame and the reference background to determine whether the camera has any interference. By adoption of the method, a plurality of camera interference conditions can be detected, such as black and white frames, the camera is blocked, the camera is moved, the reference background used in the method changes dynamically, and the impact of some natural factors is effectively handled.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method for detecting interference of a camera. Background technique [0002] In recent years, video surveillance has penetrated into various fields of national life: community security monitoring, bank ATM monitoring, traffic violations and flow monitoring, etc. Video surveillance can effectively prevent the occurrence of various criminal cases, provide timely traffic flow and road condition information, and realize accurate and rapid traffic command and dispatch. However, with the popularization of video surveillance, some suspicious persons interfere with the camera in order to avoid surveillance, such as someone covering the camera with their hands, spraying the camera with paint, turning the camera to face the wrong direction, etc., so the computer automatically and accurately Identifying these disturbances in time: that is, camera tampering detection has important practica...

AI Technical Summary

Problems solved by technology

To this end, camera interference detection needs to solve three problems: 1. The algorithm must have good real-time performance; 2. It should not be sensitive to small vibrations and rotations of the camera, but should be sensitive to obvious changes in image graphics; 3. It should be short-lived and harmless Changes cannot be detected as disturbances

[0004] There are many methods for detecting whether the camera is interfered in the prior art, but there are some problems in these methods. For example, some methods refer to a fixed background, and conditions such as lighting, rain, and scene gradients (such as day to night) will bring False detection; some methods judge whether there is occlusion by training the SVM classifier, which requires a large amount of video data collection and labeling work in the early stage, and the initial cost is high; some methods use multiple cameras to compare the scene shooting, which is not only costly, but also And there are also situations where two cameras are interfered at the same time

Images