A Method of Background Modeling for Video Images

Abstract

The invention provides a background modeling method of a video image. The method comprises: for a plurality of video image frames, each frame image is divided into a plurality of image blocks; an initial background model is established according to the first frame of the plurality of video image frames, and the initial background model is stored for each background point A corresponding sample set; for subsequent frames of the first frame, a background model for the plurality of image blocks is constructed by matching with the initial background model to form a background image. The method of the invention can quickly and accurately construct the background model.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a background modeling method of video images. Background technique [0002] With the rapid development of smart cities, the security industry market continues to grow, and the security industry will transform and upgrade to scale, automation, and intelligence. It is estimated that by 2020, the total revenue of security companies will reach about 800 billion, with an annual growth rate of more than 10%. . With the acceleration of the industrialization of artificial intelligence, by 2022, the market size of the security industry will reach nearly one trillion. Integrating computer vision technology into the video surveillance system is an inevitable trend of future development. It realizes video stream image processing, background, target analysis, etc., and builds the initial background of the model library to give the video surveillance system intelligence. Background ...

AI Technical Summary

Problems solved by technology

Taking the background modeling based on the Vibe algorithm as an example, the main problems are: 1) Ghosting problem, this is because the Vibe background modeling uses the first frame as the initial frame to initialize the background model, and there is motion in the first frame When the target is not a real background image, ghost images will appear in the detection results; 2), the static target problem, when the foreground target stays for a long time without moving (for example, a person is waiting for the subway), the moving target is gradually absorbed by the background, However, when the update speed of the Vibe background model is too fast, static or slow moving objects will be absorbed as part of the background, and the foreground objects cannot be detected correctly at this time; occluded by moving objects of human body or vehicle body), the background of the projected shadow area is falsely detected as the foreground of the moving object; 4) the incompleteness of the moving object, for example, there are a large number of holes inside the moving object, the edge of the moving object is incomplete, and there is a fault in the middle of the moving object Wait

[0004] Therefore, there is currently no background modeling method suitable for various complex occasions, and the existing technology needs to be improved to provide a background modeling with stronger robustness, better real-time performance and higher accuracy. method

Images