Video foreground and background separation method based on cascaded convolutional neural network

Abstract

The invention belongs to the field of computer vision, and provides a cascaded convolutional neural network fused with space-time clues, which is used for realizing video foreground and background separation. Therefore, the technical scheme adopted by the invention is as follows that in a video foreground and background separation method based on a cascaded convolutional neural network, two encoder-decoder type sub-networks are used to carry out video foreground and background separation; the two sub-networks are respectively an FD network for foreground detection and a BR network for background reconstruction, the FD network is used for generating a binarized foreground mask, and the BR network reconstructs a background image by using output and input video frames of the FD network; in order to introduce a space clue, three continuous video frames are taken as input; in order to improve the network applicability, the optical flow graph corresponding to the original video frame is usedas a space clue and is input into the FD network at the same time. The method is mainly applied to occasions of video foreground and background separation.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular, relates to a video foreground and background separation method based on a cascaded convolutional neural network. Background technique [0002] Foreground and background separation is a very important basic task in the field of computer vision, which has attracted more and more attention from many researchers. This technology has a wide range of applications, including motion detection, object tracking, behavior recognition, and more. Briefly, the specific task is to extract two complementary components from an input video sequence: a static background and a foreground with moving objects. Over the past decade, many approaches have been proposed to address this problem. The earliest traditional methods, such as Gaussian mixture models, nonparametric models, etc., propose to estimate each pixel independently and classify the pixels as background or foreground. The disadvantage o...

AI Technical Summary

Problems solved by technology

However, there are still some limitations to this type of work

Specifically, they only take one video frame as input, which cannot effectively capture the motion information of the foreground

In this case, the detection of moving objects in the foreground is inaccurate, causing the foreground to leak into the background

Images