Target tracking method and target tracking device

Abstract

The invention provides a target tracking method and a target tracking device. The method includes the following steps: specifying a target to be tracked in the first frame of an image frame sequence;using a local area model to determine whether there is at least one candidate local area in the current frame; if there is at least one candidate local area, carrying out primary matching on the at least one candidate local area and the corresponding local areas in the previous frame; and if determining that there is at least one deep matching area in the current frame according to the result of primary matching, performing neighborhood growing for the at least one deep matching area to obtain a target tracking result. By implementing the embodiments of the invention, target tracking for non-rigid objects can be realized under the condition of occlusion.

Description

Technical field [0001] This application generally relates to image processing, and specifically, to a target tracking method and target tracking device that are applied to mobile devices, smart TVs, and other devices, for example. Background technique [0002] In practice, there are many deformable non-rigid visual objects, and the visual appearance and shape of such visual objects may vary greatly. Therefore, especially in some complex environments, it is difficult to achieve visual tracking of such non-rigid targets. [0003] Some technical solutions have been proposed for target tracking. [0004] For example, Document 1 (CN103325126A) discloses a video target tracking method under the condition of scale change and occlusion. In order to solve the problem that the local binary pattern (LBP, Local Binary Pattern) tracking algorithm has unsatisfactory tracking effect under the condition of scale change and occlusion, a method based on the LBP algorithm and normalized moment of ine...

AI Technical Summary

Problems solved by technology

For example, Document 1 mainly focuses on video target tracking in the case of scale changes and occlusions, and cannot be used to deal with non-rigid targets; Document 2 addresses the problems of target illumination and color changes in target tracking, and improves the tracking position at the same time, which is not suitable for especially those with occlusions. Non-rigid objects; Document 3 is aimed at providing tracking speed, which is also not suitable for non-rigid targets; Document 4 is not easy to achieve a parametric shape model for any non-rigid object, especially in the case of occlusion

Images