Researches On Key Technologies Of Multiple Sensors Data Association

With the rapid development of science and technology, and the increasing complexity of variant military and industry task environment, multiple sensors data fusion technology has a great potential and a flourish prospect in a lot of application domains, such as detection, tracking, target recognition, surveillance and control and so on. Data association is a precondition for effective tracking, multi-level fusion and decision-making. The aim essence of data association is identifying the corresponding relationships among different measurements and different targets.Taking the target integration recognition for the kernel of the task, this paper tries to design data association methods in different task moments, in order to use asynchronous measurements from different sensors effectively. The aims of this paper are enhancing the quality of data association, making sure of tracking precision, and improving recognition performance.First of all, this paper proposes an effective track initiation and maintenance method using T/R-R2 multistatic radar. This method uses same source association technology to eliminate fake targets at measurement level, and gives a trick that could uncoil target locality with high accuracy and target velocity. So our method could initialize and maintain tracks effectively under long distance big error and low signal to noise ratio environment.Secondly, this paper presents a data association method which could be used for asynchronous radar and infrared sensors. This method uses combinatorial state to track targets. When the associated measurement type changes, the combinatorial state will also change according corresponding special case, and a feedback mechanism is adopted to ensure the quality of association and tracking.Thirdly, in order to conquer the deficiencies of traditional fuzzy track association method, such as low correct association rate and complexity of parameter setting, a track association method using modified fuzzy membership is proposed in this paper. This method tries to scale the similarity of tracks from a whole view, and uses a training method to get the parameters. So the practicability and correct association rate of this method is enhanced.Finally, this paper tries to take attribute information into track association method, and presents a track association method using attribute information. The method introduces a measurement based on probability to scale the similarity of two tracks, and designs a judging strategy which doesn't need to consider the complex correlation among track states and attributes. This method can increase correct association rate to a significant high level under dense target environment.The validity of the methods is validated by several groups of simulations.