Method for detecting moving object on basis of online dynamic background modeling in video

Abstract

The invention discloses a method for detecting a moving object on the basis of online dynamic background modeling in a video. The method comprises the following steps: collecting an image and carrying out pre-processing; judging whether the image frame is the first frame, if yes, initializing a background model, or else, jumping to the next step; generating a background image by utilizing a dynamic background model; extracting matching point pairs in the current frame of image and the background image; screening the accurate matching point pair and calculating a correction parameter by utilizing RANSAC; correcting the dynamic background model to the coordinate system of the current frame of image; dividing the pixels of the current frame of image into a moving pixel and a background pixel by utilizing the corrected background model; dynamically updating the background model by using the background pixel; and carrying out morphological processing, connected component analysis and area constraint on the moving pixel. According to the method, an advanced background modeling-based moving object detection algorithm under a fixed platform can be conveniently applied to an aerial photo platform, so as to detect the moving object more accurately.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method for detecting moving objects in aerial video. Background technique [0002] At present, the traditional moving target detection methods mainly include methods based on frame difference, methods based on background modeling and methods based on optical flow. For the detection of moving objects on stationary platforms, background modeling methods are most widely used due to their robustness and accuracy in extracting motion information. But for the detection of moving objects under the moving platform, the moving background increases the difficulty of background modeling, and the method of calculating dense optical flow [H.Yalcin, M.Hebert, R.Collins, and M.Black.Aflowbasedapproachtovehicledetectionandbackgroundmosaickinginairbornevideo.InProceedingsofComputerVisionandPatternRecognition, volume2 , page1202.IEEEComputerSociety, 2005] It is difficult to meet the rea...

AI Technical Summary

Problems solved by technology

This type of method requires the entire video information, so it is mostly used for offline processing, and the entire video needs to be cascaded to be corrected in the same coordinate system, which is prone to cumulative errors

[0004] Chang[Y.Chang, G.Medioni, K.Jinman, I.Cohen, Detecting motion regions in the presence of stronger parallax from moving camera by multiview geometric constraints, IEEE Transaction on Pattern Analysis and Machine Intelligence, 29(9)(2007), 1627-1641], etc. Correct the 45 images before and after the current moment in the coordinate system to the current moment image To establish a background model, if calculated one by one, the computational complexity is too high to be real-time, and if calculated in a cascaded manner, it is easy to generate cumulative errors

Images