Moving target detection method based on optical flow field clustering

Abstract

The invention discloses a moving target detection method based on optical flow field clustering. The method comprises the following steps: to begin with, generating a dense optical flow field for each frame of image of a video through a DeepFlow algorithm; then, carrying out optical flow field clustering by utilizing a DENCLUE algorithm, and clustering motion vector similar pixels into blob; then, merging the motion vector similar blob; and finally, in the time aspect, merging the adjacent multiframe blob and forming a moving target tracking path. The method has good robustness for interferences of illumination variation, shadow, noise and random swing and the like, can naturally segment the targets, the motion vectors of which are inconsistent, has a great application value for follow-up classification, tracking and retrieval and the like, is high in detection rate, and can better detect the moving targets, which can be recognized by human eyes, in the video.

Description

technical field [0001] The invention belongs to the technical field of intelligent video monitoring, and relates to a moving target detection method based on optical flow field clustering. Background technique [0002] At present, video-based moving target detection methods can be roughly divided into the following categories: [0003] A method based on background modeling. For example, the patent [200910077433.6], the disadvantage of this method is mainly that it cannot overcome the difficulties such as partial illumination and deep shadow misdetection, and it is prone to adhesion for objects with a short distance, which will cause great damage to subsequent tracking and classification. interference. [0004] Machine Learning Based Object Detector Approach. For example, the patent [201510323680.5], the disadvantage of this method is that the effect of the detector is completely dependent on the training samples, and because the target of the actual scene is ever-changing...

AI Technical Summary

Problems solved by technology

Most of the existing methods use the optical flow field as a supplement to the background modeling or target detector, such as the patent [200910236053.2], which does not make full use of the effective information of the optical flow. The main reason is that the existing optical flow algorithm The accuracy is not high. On the other hand, the time-consuming calculation limits the further modeling of the optical flow field

Images