Research On Key Technologies For Data Processing Of Target Tracking System

Information fusion or data processing is the process of integration of multiple data and knowledge representing the same real-world object into a consistent,accurate,and useful representation,which is one of the key technologies for battle early-warning system,precision strike,traffic control systems,intelligent monitoring,and etc.With the perceived environment being more complicated and various,the sensors being more abundant,and the higher requirements of sensing,the target tracking issue is getting more and more uncertain,non-linear and multi-modes.Aimed at the above issues of the modern complex environments,the research about target tracking has been carried out including non-linear estimation,data association,weak target detection and tracking,and multiple sensor information fusion.The main contributions are as follows:1.To address the inefficiency of fixed sampling number of particle filter and the difficulty in choosing the KLD sampling thresold,this paper proposes a smooth auxiliary adaptive sampling particle filter method.By designing an online metrics based particle slick,we measure the real time filtering results of the last time instant for the purpose of adaptive selection of sampling particle number of the next time instant.With smooth particles and adaptive strategy,the algorithm performs better estimation accuracy and low computational complexity.The simulation results show the effectiveness of the algorithm.2.For a certain type of drone test of maneuvering target tracking problem,a kind of interactive multiple model unscented kalman filter(IMM-UKF)is designed and compared with classical model switching method.Engineering verification results show that although both of the two methods can be applied to continuous tracking of maneuvering targets,switching method based decision logic models can iccur serious errors in model switch process,while the multiple model tracking method gives stable errors in the whole process of tracking.3.To solve the confirmation difficulties in traditional joint probability data association(JPDA)in dense clutter environment,this paper proposes a rough set based obability data association method.To avoid complex confirm matrix calculation,measurements in cross area of the tracking gate is treated differently,which is finished by dealing with the echo measurement on tracking gate based on rough set theory,and calculating of the quality of the approximated data association for target measurement classification.An adaptability test of an air defense weapons in a complex electromagnetic environment showed that compared with the JPDA algorithm,the proposed data rough set can effectively reduce the amount of calculation to achieve the same estimation precision,4.For weak target detection and tracking problem under complex environment,this paper proposes a measurement space adaptive partitioning of weak target detection before tracking method.Two layers of partitioning strategy are applied,first coarse partition of the measurement space,through measurement likelihood ratio to test the possible existing area of the targets;then fine division of target field,by designing a kind of target visibility and relevant instructions based Rao-Blackwellised particle filter for target detection and tracking,in order to adjust adaptively measuring space partition granularity.Simulation results show that the proposed method can effectively improve the target detection and tracking performance with low computational complexity.5.To deal with the possible conflicts in multiple sensor targets recognition,a local conflict assignment strategy based proof integration principle is proposed.Considering the recognition conflicts within a sensor for multiple targests as well as the recognition conflicts within multiple targets for a common target,local assignment of the conflicts and proof integration are processed.This method is verified through joint recognition examples of multiple sensors.The calculation results show this combine rule has good consistency and can effectively improve the performance of target recognition while reducing decision risk.6.For multiple targets tracking with low detect probability,in the presence of intensive clutter,a Rao_Blackwellized based particle filter to combine multiple sensors and multiple targets tracking method is proposed.By establishing Markov model of the target number and the association hypothesis Bayes full probability model of multiple sensors and multiple targets,Rao_Blackwellized particle filter could function jointly.Simulation result shows that the proposed method is able to address the non-linearity,manuvouring targets detection,association and tracking under the circumstance of unknown number of targets.7.Usually radar is firstly applied in engineering for long-range target detection and tracking,and then the tracking system will guide the infrad to track the target when the target is within the infrared field of view.However,since the infrared goniometer can only output the angle information about the target,auxiliary means to obtain the target distance information is needed in order to fully complete the tracking task.Therefore,joint radar and infrared tracking methods are studied for the purpose of designing a radar and infrared joint tracker.Performance of the fusion system is given through simulation.