Multi-sensor Data Fusion Algorithm For Target Tracking System

Recent advances in sensor technology and distributed computational algorithms, together with demands for escalating operational requirements, have contributed to the growing desire to deploy multiple sensors in various systems. By employing a syner-gistic process of fusing data from a suite of sensors (usually heterogeneous), multisen-sor systems can obtain a more complete description of the environment than is possible from an individual sensor. Single-sensor systems in general provide only partial infor-mation of the environment and rely heavily on different process strategies to extract information. These single-sensor algorithm-based systems are found wanting in their ability to resolve ambiguities, discern errors, or ascertain the cause of errors in uncer-tain environment. Multisensor systems utilize data fusion as a technique for combining information from different sources to reduce the disadvantages of the single-sensor systems. Generally multisensor systems offer greater reliability, robustness, resolution, while being more resourceful for automation.In this paper, an adaptive neural-fuzzy-based multisensor data fusion architecture (ANF-MDFA) for target tracking systems is presented. In this architecture, neural networks are employed to detect and estimate target maneuvers, and adap-tive-network-based fuzzy inference systems (ANFIS) are used to adjust the measure-ment noise covariance matrices. They are combined as an adaptive mechanism to coo-perate with Kalman filters to process measurements from multiple sensors, whose outputs are fused by a specific neural network to obtain optimal results. The results of simulation demonstrate this architecture can adjust system parameters and respond quickly to avoid miss-tracking effectively.