Research On Object Tracking Algorithm Based On Visual Memory Mechanism

Target tracking technology makes modern life fast and intelligent,but there are some limitations in the practical application affected by the environment.Using the human visual memory function for reference,the moving target can be identified accurately and tracked robustly.The tracking algorithm structure is improved to improve the tracking accuracy in complex environment.The main contents include:In order to solve the limitation of traditional tracking algorithm due to environmental factors,a cognitive model of visual information processing is established on the basis of existing memory models by analyzing and processing the visual information in the memory mechanism of human brain in complex tracking environment.Multi-feature fusion is carried out in the three-level visual memory information processing model when the moving object is modeled.HSV color histogram and FAST corner were used as the global and local features of the target for linear fusion,and the reliability of the feature model was improved by increasing the variety of target features.Experiments show that the multi-feature fusion motion model can better adapt to the change of the target and the environmental impact.For nonlinear and non-Gaussian tracking systems,particle weights and distribution states can be significantly improved by using a particle filter tracking algorithm framework with strong applicability,Grey Wolf Optimization(GWO)algorithm and particle weight adaptive adjustment.In the face of the challenges of target occlusion,transient disappearance and morphological change,template and new strategy based on visual memory mechanism were used to select the previous target template and save it into memory space according to the updating rules.In order to reduce the matching time and error,similar templates can be selected in memory space to update the target template.The average tracking accuracy of the target tracking algorithm based on visual memory mechanism in the OTB100 data set is 0.823,and the success rate is 0.793,which effectively improves the accuracy and stability of the algorithm,and verifies the effectiveness of the algorithm.Figure 37;Table 3;Reference 55...