The Research And Implementation Of Target Tracking Technology Based On Context Model

With the maturity of computer vision technology,object tracking technology has gradually attracted people's attention and gradually became a subject branch with good prospects.With the efforts of researchers from various countries,some high-level visual tracking methods have been born one after another in recent years.However,these methods often cannot deal with the adverse effects caused by complex backgrounds and have no ability to achieve high-precision,robust tracking results.Therefore,it is especially important to develop a robust tracking method.Taking into account the importance of target background information,An object tracking method based on context model has been proposed in this paper.It has optimized core algorithm,and the corresponding MATLAB GUI Visualization interface has been designed,so we can achieve the entire tracking process more conveniently.The main content of this article is expressed as:(1)With the bad influence of Fourier transform,a filter-weighting filter function is added before the tracking process to suppress the high-frequency information of the image and reduce the frequency-domain leakage effect.(2)With the problem of mutual restriction between position estimation and scale estimation in traditional tracking methods,a target tracking method which position estimation and scale estimation do not affect each other is proposed in this paper: Firstly,the position is estimated by the temporal and spatial context core framework,and then the basic framework of frequency domain filtering is used to estimate the target scale.In the process of position estimation,the proposed two interference criteria are used to determine the degree of interference of the target,and a new model updating strategy is proposed according to these two standards.The adaptive model is implemented by using the penalty parameters of the learning rate under different interference conditions.The scale estimation part creates a scale filter based on the position information obtained in the above steps,and constructs a pyramid model,accelerates it with a strategy such as sub-grid difference,and uses the posterior probability of the scale score to determine thefinal scale.(3)Subsequently,the design of the tracking method was verified experimentally,and the effectiveness of the tracking method was verified by the simulation results of different scenarios and the experimental comparison of multiple algorithms.According to the overall idea of this paper,a visual tracking interface based on MATLAB GUI is designed.It can display the important parameters such as the tracking process,position and size in real time,which is convenient for the practical application of the method.