Research On Scale Adaptive Target Tracking Algorithm

Video target tracking,as a hot research topic in the field of machine vision,has been widely used in drone reconnaissance,public security,intelligent monitoring,and missile guidance.In the past few decades,experts and scholars have proposed many excellent theories and algorithms for target tracking,but in practical scene applications,due to the interference of factors such as changes in the target scale,the algorithms still face many challenges,each algorithm still faces many challenges.In this paper,research is carried out on the basis of particle filter and correlation filter.we propose the target scale estimation method based on particle filtering,a scale adaptive correlation filtering algorithm for dual background perception and a scale adaptive tracking algorithm based on the fusion of multi-time-space perception correlation filter.The specific content is as follows:(1)For the failure tracking caused by target scale change in visual tracking,we adopt particle filter to track the target scale,and propose an algorithm of target scale estimation based on particle filter.Firstly,to enhance the discrimination of the same target at different candidate scales,we construct a probability distribution feature based on directional gradient feature.Secondly,we use prior knowledge to optimize the suggested density distribution function in the particle filter algorithm in order to improve the accuracy of scale estimation.Finally,we use reasonable scale constraint function for the second optimization of the estimated target scale.In addition,in order to reduce the interference caused by large-scale scaling,we set an appropriate threshold to realize the adaptive updating of the scale of the kernel correlation filter when we use the kernel correlation filter to predict the position,which effectively improves the accuracy of the position prediction.The experiment result shows that the proposed scale estimation algorithm is better than several typical scale estimation algorithms in terms of scale estimation and real-time.(2)In order to improve the accuracy of discriminative correlation filters on target position and scale estimation in similar background environments,this paper proposes a scale adaptive correlation filtering algorithm for dual background perception.The algorithm models the target and the background information according to the background-aware correlation filter algorithm.On this basis,the background information around the target is extracted as a negative sample and added to the learning of the tracker,so as to suppress the background information,increase the distinction between the target and the background,and further improve the accuracy of the algorithm in target detection.(3)This paper proposes a scale adaptive tracking algorithm based on the fusion of multi-time-space perception correlation filter,which combines the time-space characteristics of target and background,to improve the reliability of video target tracking under different complex environments.Based on the correlation filter,the time characteristic is established by calculating the consistency of target and filter changes between frames;Based on the neighborhood information of the tracking target,the mask matrix is introduced to extract the spatial information of the target;Finally,in the objective function,the time-space perception constraints are constructed to enhance the learning ability of the correlation filter to the time-space binary information,thus enhancing the robustness of the filter to the interference information.In order to improve the ability to adapt to different complex environments,an independent time-space perception correlation filter is established in the feature space of video color and orientation gradient,and an adaptive fusion mechanism of the tracking results of the two is established to accurately calculate the position and scale of the target,thereby effectively improve the generalization ability of the algorithm for different complex environments.