Research On Application Of CKF Algorithm For Target Tracking

Nowadays,tracking is a hot technology,is widely used in all areas and sectors. In the field of video surveillance, accurate target tracking helps technicians quickly target behavior;in the military field, precise target tracking to quickly lock radar targets, so as to accurately predict actions do next. Target tracking in the field, research on filtering algorithm and its application to improvement of tracking performance plays a crucial role. When the target State changed abruptly, traditional tracking algorithms can lose track of the target. Introducing appropriate filtering can significantly improve the accuracy of tracking. Based on this, this article combines filtering and tracking algorithm, used in video object tracking and radar target tracking in the two-dimensional space of two cases, allowing new algorithms to meet accurately tracking requirements.This paper outlines the development of target tracking and filtering, clear tracking of core issues. Model for establishing a target tracking,and target tracking based on Bayesian filter theory principles and process simulation.Paper begins with the linear Kalman filtering, analyzing in three non-linear filtering algorithm EKF,UKF and CKF, and equivalent flops to the precision and complexity,summarizing the models and the advantages and disadvantages of various algorithms. Current popular UKF and CKF non-linear algorithms on different dimensions of model simulations,compare the mean square error, and determines when the dimension the system state greater than three, selecting the need for CKF algorithm.Combine CKF through Mean Shift, by CKF algorithms constantly predict the next frame is to update the target point location, so as to make up for the deficiency of Mean Shift algorithm,state control of new algorithms in a car video, change the state of motion of the car.By analyzing and comparing the two methods of tracking accuracy, the new algorithm for tracking performance.The traditional CKF algorithm is improved, combined with MFNN and adaptive filtering for noise statistics of the model is the NACKF method under uncertainty,and combined with IMM algorithm applied to a state change in the two-dimensional space of particle tracking, and finally through the simulation of new algorithm on tracking accuracy.