The Track Algorithm Research Of Multi-Sensor Data Fusion

Multisensor data fusion is a rising discipline, and has wide applicationsin areas of radar, sonar, navigation, automatic control, communication and robot.Data fusion techniques combine data from multiple sensors or multiple kinds ofsensors, and related information from associated databases, to achieve a betterperformance of detecting, tracking, classifying targets than performance couldbe achieved by the use of a single sensor alone. The distributed data fusioncan rise the system's reliability and survivability effectively. This thesisconcerns the problem of multi-target tracking by data fusion technology. Themain contributions of this dissertation are summarized as follows:1. First, this paper introduces definition, fundamental principle, andapplications of multi-sensor information fusion. Then it carries on the summaryto the multi-sensor information fusion function model and its the multi-sensorintegration and the data fusion significance.2. The basic principle and type of maneuvering target tracking is introducedin the dissertation. It lays the foundation for the behind content furtherdiscussion.3.The trend of target tracking technique is forming multi-sensor data fusionsystem. Essentials of target tracking such as coordinate system choice and targetmaneuvering model setup are discussed in detail. Several algorithms on trackingassociation are investigated by pointing out their merits and defects. The stricttheory analysis shows that the PDA algorithm is applied.4. According to the ideas presented by Chen J. and Leung H that the motiondirection of the target is also a key factor in data association, adirection-tracking gate based on the pseudodirection concept is ameliorated.By combining the ellipse tracking gate with the direction tracking gate, a newtwo-gate tracking algorithm is obtained, which can decrease the number offeasible association events and the load of JPDA calculation. At last, computersimulations are performed to demonstrate the effectiveness of the new algorithm.5.A new multi-sensor multi-target tracking algorithms is verified for fastcomputing in real time tracking. The theory analysis shows that its estimatefar is smaller than the original order and the parallel multi-sensor jointprobability data association algorithm. The simulation results show that it cantrack several targets simultaneously with high accuracy.