Research On Image Tracking Algorithm Based On Object Feature And Spatial Information

Object tracking is an important branch of computer vision.It has a wide range of applications in human-computer interaction,intelligent monitoring,military,and3 D reconstruction.It has important research value.object tracking core.Is to correctly detect the target,with a reasonable feature to express the target,and then use the algorithm to track the object.In the video object tracking,due to the complexity of the target itself and the background,there is no uniform algorithm to deal with various situations.Mean Shift and particle filter algorithms are commonly used in object tracking algorithms.Considering the real-time,this paper takes the Mean Shift algorithm as the core,analyzes the advantages and disadvantages of the algorithm,and puts forward the Cam Shift algorithm in the case of light,and compares the two methods.These two methods do not take into account the impact of background factors,so that for the complex background of the case there will be tracking the target deviation.Aiming at the tracking in complex background,the CBWH background optimization algorithm is proposed and the same as the tracking effect of the BWH algorithm and the Mean Shift algorithm is verified.A background model is established,using the background model to obtain a weight transformation,the purpose of the transformation is to make the target occupies a higher proportion of the characteristics of the more prominent,and similar to the background of the specific gravity decreased,thus reducing the background for the target Interference,making CBWH more accurate for tracking in complex contexts.When tracking the target object with dramatic changes in the target,only one feature is used to describe the target,and the final tracking target deviation can also be caused.Therefore,this paper uses the spatial characteristics and color characteristics of the target to describe the target model,and the CBWH background optimization algorithm is used to track the target.Through experiments,it is proved that the object can be well tracked when the color of the target object changes drastically.