Research And Realization Of Object Trakcing Method In Complex Scene Based On Correlation Filtering

Object tracking is an important research branch in the field of computer vision,which is widely used in intelligent security and monitoring,human-computer interaction,intelligent transportation.During actual tracking,object is easily affected by factors,such as illumination variation,scale changes,surrounding background information changes,rotation,motion blur,fast motion,resulting in inaccurate tracking results.How to robust track object is one of the difficulties in the field of object tracking.After the theory of correlation filtering is applied in the field of object tracking,it has obtained better tracking effect.While improving the accuracy of object tracking,the speed of algorithm processing has been improved to a certain extent.However,the correlation filtering algorithm uses a fixed model update mechanism to update the filter model and the object feature vector,resulting in the tracking model not being well adapted to tracking in complex scenes.Therefore,this paper focuses on researching on the correlation filter tracking framework with object relocation mechanism to solve the problem of inaccurate object tracking results in complex scenarios.The specific research work is as follows:(1)During tracking,in order to adapt to the status of object movement and changes in surrounding environment well in complex scenes,and make the model update through adaptive learning.This paper improves the adaptive learning model update mechanism based on the existing tracking framework.Model maintains the discrimination and tracking performance of object in complex scenarios through update learning rate adaptively.(2)At the same time,because the difference of the complex scene,based on the existing tracking framework,the detection threshold is dynamically adjusted to increase the extensive of determining whether the tracking result is reliable.By conducting experiments on the recognized and authoritative experimental data sets OTB-2013 and OTB-2015 in the field of object tracking,the results show that the combination of the above two improvements can better tracking object in complex scenes.Compared with other methods,it has good tracking accuracy and success rate.(3)Object tracking system was designed and implemented using Matlab GUI technology.Tests show that the system can track object in complex scenes in the experimental video set.