Target Tracking And Attitude Estimation Based On Correlation Filtering And NCC And Its Application In Augmented Reality

Based on the correlation filtering and template matching technique(Normalized cross-correlation),this paper makes innovative research and improvement on the tracking technology in AR(Augment Reality)application,so that the target tracking algorithm in AR can be real-time under 1080P/25HZ conditions.Calculate the position information and direction information of the 3D object to assist the actual AR display.In this paper,an AR tracking algorithm based on template matching technology is designed.The online feature construction,uniform feature selection and target matching secondary search algorithm are designed and implemented.The AR tracking algorithm based on template matching can track target in real time for 1080P video image tracking speed between 25 and 30 fps,and can well cope with target occlusion.However,when the algorithm changes in the target scale and the target moves quickly,the tracking success rate will drop rapidly.The distance between the target and the lens is between 0.3 m and 1 m to ensure high-precision tracking.The success rate of fast target movement is less than 20%.Aiming at the scale change problem,this paper designs the knot and correlation filtering and the target search algorithm of multi-layer offline pyramid,so that the target distance lens can guarantee high-precision tracking at 0.1 m to 2 m,and the tracking speed of 1080P image is increased to 30 fps.Between 45fps.Aiming at the fast motion problem of the target,this paper designs a target position prediction algorithm combined with Kalman filter to make the target's tracking success rate in fast motion reach 65%.The target tracking algorithm in the final AR can cope with target occlusion,scale change and target fast motion under the condition of ensuring real-time tracking.