A fast moving object tracking method based on KCF hybrid MFO

Abstract

The invention provides a fast moving target tracking method based on KCF mixed MFO, which comprises the following steps: initializing target state parameters and optimizing model parameters; The KCF tracking method is executed to calculate the maximum response value of the target in the first few frames, and the confidence threshold initial value is obtained by the KCF tracking method. According to the relationship between the maximum response value of the current frame and the confidence threshold, different methods of generating the base sample images are determined. When the maximum response value is higher than the confidence threshold, the base image samples are selected randomly and the KCF method is used to track the target. When the maximum response value is lower than the confidence threshold, the MFO searching mechanism is used to search the global optimal solution and as a new base image sample, and then the KCF method is used to track the target. Dynamically updating confidence threshold, repeating the above operation to achieve effective tracking of fast moving objects. The invention can effectively realize persistent tracking to the fast moving target in the video, and improves the adaptability of the tracking method under the complex scene.

Description

technical field [0001] The present invention relates to the technical field of target tracking, in particular to a fast-moving target tracking method based on KCF hybrid MFO, which realizes the continuous tracking of the target, and also includes the state space search under the condition that the target produces fast motion between adjacent image frames of the video Mechanism and target continuous tracking capability. Background technique [0002] The visual tracking problem focuses on predicting the trajectory of moving objects in a video. Currently, visual tracking has been applied in many practical applications, such as robotic services, self-driving cars, surveillance, human-computer interaction, and more. Visual tracking is an open and challenging problem, where a tracker must be able to track target objects over long periods of time, even in complex scenarios such as drifting objects and background occlusions. Given the initial position of the target object in the i...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the technical problems of low operating efficiency and poor versatility of existing moving object tracking methods, the present invention proposes a fast moving object tracking method based on KCF hybrid MFO, which introduces the moth-flame (MFO) search strategy into the kernel correlation filter In the tracker (KCF) design, it is based on the combination of traditional local search algorithm and optimized global search algorithm, and uses dynamic threshold judgment to re-determine the target, which improves the operating efficiency and can adapt to fast-moving video target tracking

Images