Research And Implementation Of Planar Object Tracking Algorithm Based On Convolutional Neural Network

Visual target tracking is a particularly important branch in the field of computer vision.As one of the categories of visual target tracking research,planar object tracking is not only a hot field of computer vision academic research,but also augmented reality,surveillance,robot vision and other directions.A wide range of applications.In recent decades,more algorithms have emerged to solve the problem of planar object tracking.At present,most algorithms belong to the traditional planar object tracking algorithm.The traditional planar object tracking can be reduced to the nonlinear optimization problem.The core goal of the algorithm is to minimize the objective function.Since the problem itself is a non-convex problem,the solution process is It is easy to get a local extremum solution and the tracking effect is not ideal.At present,the image processing tasks based on deep learning networks have attracted much attention.In view of the fact that there are few related studies on the combination of deep learning and planar object tracking,this paper proposes to apply the deep learning network to the planar object tracking task and explore A more accurate and efficient algorithm for solving planar object tracking problems.The main research work of this paper is as follows:(1)Investigation and analysis of the problem.In this paper,a large number of existing planar object tracking algorithms are investigated,and the local extremum problems existing in the existing algorithms are analyzed.(2)A planar object tracking algorithm based on convolutional neural network is proposed.The proposed algorithm uses the convolutional neural network architecture to realize the end-to-end tracking training process.The overall structure simulates the three stages of feature extraction,matching and evaluation of transformation parameters in the traditional planar object tracking process.Compared with traditional algorithms,this algorithm can capture the correlation between image features and improve tracking performance.(3)Verify the proposed algorithm.In the Linux system,the planar object tracking algorithm based on convolutional neural network is implemented based on Python language and Pytorch deep learning framework.The proposed algorithm is validated by the challenging TMT and UCSB standard data sets.The experimental results fully demonstrate the effectiveness of the planar object tracking algorithm based on convolutional neural network.At the same time,the tracking accuracy of the proposed index shows that the proposed planar object tracking algorithm has better tracking performance.