Research On Target Tracking Algorithm In Video Sequence

With the continuous development of computer technology,computer vision has become a hot research field.Among them,target tracking is a key technology in many research areas of computer vision,which has wide application value in production.In recent years,target tracking algorithms have achieved good results,such as correlation filter tracking algorithms with high processing speed and high accuracy.Tracking algorithms have higher running speed and better tracking accuracy,but these are still not effective in tracking scenarios such as occlusion,fast motion and scale variation.In order to solve the above problems,the improved kernelized correlation filters is used to realize target tracking in complex scenes.It is also applied to ship recognition to realize the automatic recognition and tracking of ships.Details of the work are as follows:In the research of tracking algorithm of multi feature adaptive learning mechanism,this paper integrates multiple features by dynamically calculating the weight.This can solve the problem of insufficient expression ability of single hog feature in motion blur,deformation and other scenes.The traditional tracking algorithm uses fixed scale to track the target.In this paper,the pool of target scale is established and the scale filter is trained to realize the scale recognition.In view of the poor tracking effect of the fixed learning rate in the scene of illumination variation,occlusion,or fast movement,this paper combines the moving speed of the target with the average peak-to correlation energy to dynamically adjust the learning rate of the filter.In the research of the detection of reverse scale and relocation algorithm,the reverse scale detection mechanism is used to improve the accuracy of scale filter for small-scale target recognition.For the problem that the target is blocked or moves out of the detection area due to rapid movement,this paper dynamically starts the target detection according to the detection factor.For the application of target tracking algorithm in ship recognition,the moving target is detected by frame difference method.In view of the influence of watermark tailing on target recognition and tracking,this paper uses the designed watermark tailing classifier to eliminate it.In the end,the ship classifier is designed and combined with the tracking algorithm to realize the automatic tracking of ship objects.Finally,experiments are carried out and compared with common mainstream tracking algorithms.The results show that the algorithm of this paper has good tracking success rate and accuracy.It can also meet the requirements of real-time tracking.In the intelligent ship identification and tracking experiment,the ship identification and tracking algorithm in this paper has good stability and accuracy.It also meets the requirements of performance and function.