Method and system for extracting foreground target by removing ghosting

Abstract

The invention discloses a method and system for extracting a foreground target by removing ghosting, and belongs to the technical field of foreground target detection. The method comprises the following steps: acquiring video stream information, selecting pixel values of a plurality of arbitrary position points of an initial frame image in the video stream information as a sample library, selecting a plurality of pixel values in an arbitrary position neighborhood of the initial frame image with equal probability, and assigning values to the sample library to generate a background model; selecting any frame in the video stream information, and carrying out foreground and background classification on any pixel point of the any frame; updating the pixel value of any pixel point into a samplelibrary of the background model according to a preset probability; and determining any pixel point as a ghost image pixel point, and removing the ghost image pixel point to extract the foreground target. According to the method, the dynamic model and the flicker degree are introduced to evaluate the dynamic degree of the pixel points, the sampling distance threshold value and the matching threshold value are updated in a self-adaptive mode, the foreground extraction accuracy is improved, and meanwhile the omission ratio is reduced.

Description

technical field [0001] The present invention relates to the technical field of foreground object detection, and more particularly, to a method and system for extracting foreground objects by removing ghost images. Background technique [0002] Foreground object detection is to automatically segment the video sequence of the camera into interesting foreground objects and backgrounds, and the results are the basis for subsequent research such as object tracking, counting, recognition, and classification. How to deal with the variability of the scene (such as dynamic background, shadows, etc.) is the biggest challenge this technology faces at this stage. [0003] The current mainstream foreground object detection methods include frame difference method, optical flow method and background subtraction method. The comprehensive advantages in real-time and accuracy make the background subtraction method a current research hotspot. The core task of this type of method is to establi...

AI Technical Summary

Problems solved by technology

[0005] 1) The strategy of classifying and updating with fixed values ​​is difficult to adapt to dynamic backgrounds, such as water flow, shaking branches and leaves, etc.;

[0006] 2) If there is a foreground target in the single frame used to initialize the background, ghost images will appear in subsequent detections

[0007] The mixed Gaussian model has high computational complexity, poor real-time performance, and it is difficult to eliminate false targets caused by dynamic backgrounds. The representatives of parameter background models are kernel density estimation (Kernel Density Estimation, KDE), codebook model (CodeBook) and vision Background extraction (ViBe), the first-in-first-out update strategy for observations makes this method unable to adapt to long-term events. CodeBook only relies on images with a specific number of frames to train the background model, which is difficult to adapt to illumination changes and irregular backgrounds. Complicated situations such as sports

Images