Research On Target Tracking Algorithm Based On Human Brain Incentive-Feedback Mechanism

Based on the principle of visual stimulation mechanism of human brain,the target tracking algorithm is designed by combining the deep neural network,visual memory mechanism and attention selection mechanism.Through extracting depth features and training adaptive filters,the target can be tracked accurately and in real time.The research results are as follows:Through the research and analysis of the memory mechanism and attention mechanism in biological vision incentive mechanism,a model framework is established,and the target tracking algorithm is designed by combining deep learning feature extraction and correlation filtering fast search algorithm.Aiming at the challenges of light change,scale change or occlusion faced by target tracking algorithms,a correlation filtering algorithm based on ResNet feature extraction is proposed,which combines visual short-term memory mechanism for fast tracking and result correction.With the adaptive update module of correlation filter,a robust and stable classifier is trained,which shows the high accuracy and success rate of depth residual network in target tracking,and verifies the overall effectiveness of the algorithm.In view of the trend that the running speed and tracking success rate of traditional target tracking algorithms decrease when facing complex situations,a visual target attention mechanism tracking algorithm based on MDNet is proposed.Based on the advantages of multi-domain network in tracking field,the positive and negative samples are classified and concentrated,and the attention weights of both samples are adjusted.While ensuring the high tracking accuracy of MDNet,it also maintains a certain real-time performance.The tracking algorithm based on human brain incentive mechanism has realized the application of human brain vision mechanism in computer vision to a certain extent.By introducing biological mechanism into machine learning,it has achieved practical effectiveness in target tracking task,which provides some ideas for the research of computer vision and machine vision related fields.Figure 43;Table;4 Reference 55...