ViBe-oriented foreground ghosting removal method

Abstract

Faced by the problems that in the foreground extraction of a visual background extractor (ViBe), ghosting arises easily and that when an object with dense foreground passes a ghosting area, it requires a long time to eliminate the ghost, the invention provides a ViBe-oriented foreground ghosting removal method in order to suppress the generation of ghosting and remove the ghosting rapidly in the foreground extraction. According to the invention, a second frame detects the number and the positions of the ghosting generated in the model initializing process, on the basis of which, an intersection set between the background area of each frame after the second frame and the detected background area is obtained. The obtained set represents the background points blocked by the foreground of the first frame but not blocked by the current frame. The re-initialization operations are made to the local model of the pixels in the set and after the operations, the set of the ghosting areas is updated until the ghosting is completely removed. The method of the invention can remarkably suppress and remove ghosting generated from the foreground extraction by a visual background extractor.

Description

technical field [0001] The invention relates to the field of computer vision technology and video image processing, in particular to the foreground object detection and ghost suppression technology oriented to ViBe (Visual Background Extractor, i.e. a visual background extractor). Background technique [0002] With the popularization of high-definition cameras and the rapid development of the Internet of Things, video surveillance is widely used in people's production and life, resulting in massive video data. Traditional manual inspection becomes more and more difficult with the increase of video data. , how to use computers to extract useful information from massive videos has become one of the research hotspots in the field of computer vision. The first step in extracting meaningful information from a video is to detect and extract moving objects from the video. Currently, there are three commonly used methods for moving object detection: optical flow method, inter-frame ...

AI Technical Summary

Problems solved by technology

However, since ViBe only needs the first frame to complete the initialization of the background model, and only relying on the first frame does not contain spatio-temporal information, there is a disadvantage that it is easy to introduce ghost images, and when foreground objects continue to pass through the ghost area, ghosts will appear. The shadow is continuously strengthened and cannot be eliminated

This will seriously affect the accuracy of background modeling, thus affecting the effect of ViBe in foreground detection applications

[0004] In view of this shortcoming, the common methods for suppressing ghosting are: 1. Preprocess the first n frames of the video, and use the frame difference method to first obtain the background image that does not contain the foreground object. After initializing the model, the ghosting can be effectively eliminated. However, due to reserving n frames for preprocessing, it also prolongs the model initialization time. Compared with the original ViBe, only the first frame is required to complete the initialization. This method achieves a certain degree of ghost elimination, but However, the real-time performance of the method is sacrificed; 2. The ghost suppression method based on the threshold has a certain suppression effect on the ghost, but because the setting of the threshold makes the method have a certain delay, the elimination speed of the ghost still has a certain effect. room for improvement

Images