Research And Implementation Of Object Tracking Based On Optical Flow

Object tracking is one of the core tasks in the field of computer vision.Given a sequence of continuous video frames and the spatial position of the target to be tracked in the first frame,the object tracking requires the model to find the position of the target in all frames.Object tracking not only needs to classify and locate the target,but also needs to analysis the timing context information of each frame,and need multiple technologies in the field of computer vision.Although the field of object tracking has been extensively studied in recent years,it is still a challenging problem due to the difficulty and complexity of the tracking task itself.This paper proposes a object tracking method based on optical flow field,which estimate the motion state of the target including displacement,scaling and rotation.Since the position of the target in the first frame is known,once the motion state of the target in each frame is estimated,the position of the target in each frame can be known.To achieve this goal,the proposed method uses optical flow to model the motion state of the target.The method first models the optical flow as a two-dimensional optical flow field,and innovatively uses the vector differential operator to obtain the motion state of all pixels in the frame,and then obtains the motion state of the target rectangular frame through aggregation.In order to further utilize the appearance characteristics of the target and improve the accuracy of object tracking,this method uses spatial attention mechanism and feature fusion technology to further improve the expressive ability of the model.This paper evaluates the algorithm proposed in this paper on two mainstream object tracking benchmarks.The evalutaion results show that the effect of the object tracking method based on the optical flow field surpasses the baseline model.Based on the object tracking algorithm proposed in this paper,this paper designs and implements a object tracking system based on optical flow.The system takes the optical flow-based object tracking algorithm as the core,which can read video data from multiple types of data sources and track any object in the video.This article designs a comprehensive test case to test the function of the system.The test results show that the system implemented in this paper can complete all the functions of the predetermined design.