Research On Airborne Multi-Sensor-Multi-Target Data Fusion

There are many research aspects in airborne multi-sensor information fusion system. This dissertation focuses on the practical problems of engineering applications, such as airborne multi-sensor detecting modeling, target recognition, target state estimation and target tracking as well. The main contents and results in this dissertation are as follows.The detecting and tracking models for multi-sensor system based on liner programming are well studied and established, aiming at the practical airborne fire control and guide systems. The value of effectiveness for liner programming is the ratio of the target's relative value to the mean time of tracking, that is, the overall value of the sensors locked in each time unit is served as the objective function. The liner programming is practically feasible since the establishment of target priority and mean time for tracking. Meanwhile, the coordinated detecting and fusing model for Radar and IRST is also given.Multi-hypothesis tracking algorithm was primitively applied for the data processing of infrared image-tracking system, and it was applied for the target-tracking of single-sensor at all times. Nowadays, there are some large-scaled and advanced detecting systems overseas which adopt MHT algorithm; however, there is no technical detail. In this paper, a specific and efficient MHT algorithm for long-scan-period-radar and centralized fusion structure is established with engineering realization, based on the practical military equipments and the application circumstance.For target recognition, evidence combination equation is applied with appropriate revise. The application of evidence theory in multi-sensor-multi-measurement airborne radar recognition is quite effective for bettering the single sensor or single characteristic recognition.The image tracking and angle only tracking techniques for long distance IRST are also investigated. For the long distance tracking of single small target by infrared image sensor, as long as the tracker could offer feedback information, i.e. the predicted location and distribution of target, the sensor would use Bayes detection. As a result, a variable greyscale threshold is implemented in image filtering, reducing the number of candidate measurements, which is in favor of the PDAF-based tracking method. As for MHT application, image overlapping is prevalent when targets are far from sensor and the targets density is relatively high, even if the false measurement density is not so serious. To cope with the image overlapping, a multi-assignment method is established based on one-to-one two-dimensional assignment. The fused tracking algorithm for multi-angle-only-sensor system is studied, which is a weakness in current information-fusion research. First, the CRLB is used to analyze the tracking precision by two sensors from different platforms, considering bandwidth influence. Second, the constant velocity model is applied to the EKF tracking in a single site manner. Calculation and simulation validates the significance of two passive sensors tracking. The applicability of MGEKF in multi-angle-only-sensor case is investigated and proved. Precision analysis is given to discuss the combination orientating problem between infrared and radar system.