Research Of Data Fusion Algorithm For Multiple Sensor Target Tracking

As a new and developing crossed subject, Multi-sensor data fusion technology has a board prospect of application in many domains. Multi-sensor target tracking is a paradigm of data fusion technique with target tracking problem, which can be used to increase the precision of target state estimation in this technology. It has more advantages than the case of single sensor. Data fusion algorithm is one focus of study for solving target tracking problem.The thesis researched track fusion algorithms and filter estimation algorithms of the target tracking, which was improved, and simulated the experiment for the track fusion algorithms with the different filter estimations. The design technique and applications of multi-sensor data fusion system were discussed in the thesis. The main work of this thesis was listed as followed:(1) The thesis analyzed and compared extended Kalman filter, unscented Kalman filter and particle filter method; extended Kalman filter and unscented Kalman filter was used respectively as the proposal distribution of particle filter, the current moment of measurement was given with full consideration, that making the particles distribution closer to the status of the posterior probability distribution, that is the improved particle filter; the thesis analyzed interactive multiple model estimation and simulated the experiment. The result shows that the algorithm was effective for multiple model target tracking.(2) The related techniques of multi sensor data fusion were discussed. The thesis analyzed the weak point of the simple fusion algorithm, the co-covariance fusion algorithm and adaptive track fusion algorithm. Those algorithms were unable to meet the uniformity of co-covariance information. The thesis used a data fusion mechanism, which didn't need independence assumption. That is called covariance intersection algorithm. Because of its joint covariance information associated with independent error and correlated error, the updated covariance was improved. The improved covariance intersection algorithm was formed with isolated correlated error and the independence of error components from the covariance.(3) The thesis designed a data fusion algorithm. Filter and estimation of data fusion process were used by the interactive multi-model and improved particle filter. And the improved covariance intersection was as the global integration algorithm.(4) The design technique and applications of multi-sensor data fusion system was discussed, which includes methods of designing and developing and performance evaluation. And the target tracking system has applications in engineering.In the light of some filter algorithms, the precision of split covariance intersection algorithm is higher than covariance intersection algorithm, which was proved by the certain series contrast experiments.