Target tracking method on basis of online study for context relation of space-time structure

Abstract

The invention discloses a target tracking method on the basis of online study for a context relation of a space-time structure. The method is used for tracking target images in videos, and includes steps of creating a time-structure context model and a space-structure context model for target tracking; sampling areas, where an object probably appears, in a video frame to obtain sample images; inputting the sample images into the time-structure context model and the space-structure context model to obtain time probability results and space probability results which indicate various sample points in the sample images are areas with the target; and fusing the space probability results and the time probability results to obtain posterior probabilities indicating that the various sample points belong to the areas with the target, and using a sample point corresponding to the maximum value of the posterior probabilities as a final result of target tracking. The target tracking method has the advantage that the target can be accurately positioned via automatic online study for representation features of the target under the condition that any information of the target and any information of scenes with the target are unknown.

Description

technical field [0001] The invention relates to the technical fields of pattern recognition and computer vision, in particular to target tracking technology, in particular to a target tracking method based on online learning of spatio-temporal structure context. Background technique [0002] Object tracking has many practical applications, such as human-computer interaction, video surveillance, virtual reality and so on. For the vast majority of cases, the tracker needs to track the target for a long time, and it is difficult to obtain the prior knowledge of the scene where the target appears and the target appearance information, so it is necessary for the tracker to perform online tracking according to the changes of the target during the tracking process. Automatically learns to be more robust to changes in the target. [0003] In the actual application process, the most basic and effective information of the moving target is the appearance information of time and space,...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is that the existing target tracking methods cannot effectively use time and space representation information, so that the target tracking effect is not good in non-restricted conditions.

Images