Cuckoo search and KCF fusion-based method for tracking target with sudden change motion

Abstract

The invention discloses a Cuckoo search and KCF fusion-based method for tracking a target with a sudden change motion. The method comprises the following steps of initializing state parameters of a target and optimizing initial parameters of the method; obtaining maximum response values in first multiple frames of the target by adopting a KCF tracking method, and calculating an initial value of a credibility threshold; according to a relationship between the maximum response value of the current frame and the credibility threshold, determining different basic sample image generation modes: when the maximum response value of the current frame is greater than the credibility threshold, randomly selecting a basic image sample, and executing the KCF method to track the target; and when the maximum response value of the current frame is smaller than the credibility threshold, obtaining a globally optimal target prediction state by adopting a Cuckoo search mechanism, generating a new basic image sample, and executing the KCF method to track the target; and dynamically updating the credibility threshold, and repeating the steps to realize target tracking. According to the method, continuous tracking of the target with the inter-frame sudden change motion under a dynamic camera is effectively realized; accurate tracking of the target with the inter-frame sudden change motion is realized; and the adaptability of the tracking method in a complex scene is improved.

Description

technical field [0001] The invention belongs to the technical field of target tracking, and in particular relates to a method for tracking a sudden moving target which combines Cuckoo search and KCF to realize continuous tracking of the target, and especially relates to the state space in the case of a sudden change in motion of the target between adjacent image frames of the video Search mechanism and target continuous tracking capability. Background technique [0002] Affected by factors such as the tracking environment, the target itself, and camera imaging, the object of interest often has abnormal motion mutations during the tracking process, which makes it difficult for many tracking algorithms proposed based on the assumption of motion smoothness to adapt and lead to failure. The sudden motion of the target has become one of the important factors for an algorithm to achieve continuous tracking. It is necessary to design a new method that can adapt to the sudden motion...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the technical problems of low operating efficiency and poor versatility of existing moving target tracking methods, the present invention proposes a mutation moving target tracking method that combines Cuckoo search and KCF, and introduces the cuckoo search strategy into the kernel correlation filter tracker In the (KCF) design, the method of generating the base sample image is determined according to the confidence threshold to enhance the ability to evaluate the sudden movement state; the regression problem in the spatial domain is equivalent to a correlation filtering operation, and the time-consuming convolution operation is used. The multiplication operation is replaced, which improves the operating efficiency of the algorithm, and finally realizes the video target tracking method that can adapt to sudden changes

Images