Unsupervised target dense tracking method based on video coloring

Abstract

The invention discloses an unsupervised target dense tracking method based on video coloring, and the method comprises the steps: constructing a target tracking model based on video sample frames which are arranged in a time sequence and respectively comprise a preset target object, and then completing the tracking of the preset target object through employing the target tracking model. The target tracking model comprises a feature extraction network used for obtaining feature maps corresponding to video sample frames, and a dynamic adjustment module used for outputting a feature map group formed by a preset number of feature maps in all the obtained feature maps. And the target prediction module is used for predicting the position parameter and the area parameter of the preset target object in the next frame of feature map of the feature map output by the feature extraction network in real time, and the method can improve the tracking precision of the preset target object and can adapt to the tracking scene of the violent change of the preset target object.

Description

technical field [0001] The invention relates to a dense target tracking method, in particular to an unsupervised dense target tracking method based on video coloring. Background technique [0002] The problem of target tracking is widely studied in computer vision, and the existing target tracking often requires a large number of image labels and the tracked target categories are limited, so the unsupervised target tracking problem has attracted extensive attention and in-depth research in the academic community. There are four main types of solutions to the existing unsupervised target tracking problem, namely: methods based on correspondence flow, methods based on temporal cycle consistency, methods based on multi-grid prediction filters, and methods based on video coloring. The present invention is based on the method of video coloring, the general flow of this type of method is: given a certain frame in the video, traverse each pixel point p of the frame i , find the mo...

AI Technical Summary

Problems solved by technology

[0002] The problem of target tracking is widely studied in computer vision, and the existing target tracking often requires a large number of image labels and the tracked target categories are limited, so the unsupervised target tracking problem has attracted extensive attention and in-depth research in the academic community.

Images