Research On Target Tracking Based On Sparse Subspace Convolution Neural Network

With the development of science and technology,the target tracking technology is open to the civil science and technology from the military field.At present,it is not only the research focus in the field of computer vision but also has broad prospects and huge economic value.Target tracking technology is widely used in mobile platform and robot information acquisition.Due to the different scenes of platform application,along with the complex changes of environment,the stability of the whole tracking system can be affected.Aiming at the problem of particle degradation in data processing,a particle filter algorithm based on sparse subspace is proposed.First of all,the author propose to divide the sample into positive,negative and transition samples in the improved particle filter.Then,following the human visual nervous system,the sparse subspace is introduced into the particle filter,a sparse optimization model is established and the target classification is achieved,and the center of the cluster is obtained to achieve the target tracking.The experimental results show that this algorithm has better robustness.Compared with the particle filter it can be more effective in solving particle degeneration problems.Aiming at the problem of data redundancy and processing speed of particle filter,a convolutional neural network tracking algorithm based on sparse subspace is proposed.The sparse subspace model is used to select the candidate regions with high similarity to the target state for follow-up tracking.The convolution neural network model is used to update the tracking data to reduce the influence of the target apparent change.The experiment shows that the algorithm can deal with the data redundancy in target tracking well and has faster data processing speed.A sparse subspace model is proposed,which aims at improving the accuracy of target extraction and improving the accuracy of target extraction.It provides a new idea for target tracking research.