Research On Target Tracking Theory And Method Based On Heterogeneous-Sensor Information Fusion

Since the middle of the last century, target tracking has been widely used in industrial robots, video surveillance, missile interception and other fields. With the rapid development of information technology, the multisensor information fusion has become the focus of target tracking to improve the real-time and accuracy of target tracking system. In the paper, we study the one of the key components of the new generation of aerial refueling technology: target tracking system. Under the requirements of non-contact, high precision and real-time, the problems in the processing of target tracking such as outliers, measured-correlation and time-delay are solved effectively by the complementary of multi-sensor information. In summary, the main contributions of this paper are listed as follows:(1) Under the limitation and requirement of target tracking system in air refueling,the target tracking system which comprises the binocular vision sensor and laser range finder is proposed. In order to improve the poor measurement accuracy of the binocular vision sensor along the optical axis direction, we use a one-dimensional laser range finder to measurement correction. In this system, the azimuth angle and elevation angle of the laser range sensor are influenced by the measured values of the binocular vision sensor, it leads the tracking errors increasing, we analysis the error propagation and verify the feasibility of the system.(2) After using the time registration of interpolation extrapolation algorithm, there is a large amount of information loss in the high measured-frequency sensor system.The weighted average data compression technology is used to improve the signal to noise ratio (SNR) and information utilization ratio. In order to detect, eliminate and correct the outliers, the improved Wasserstein distance is proposed. The simulations show the feasibility and accuracy of the algorithm to detect and correct the outliers.(3) According to the actual motion between two aircrafts, each stage state of the target is divided into static state, uniform motion state, decelerated motion state and disturbance state. In order to approach the actual motion state, correlation gate technique of rectangular gate is used to convert each motion state adaptively. The simulations show that the model has good adaptability to the system.(4) Aimed at the problem of measured constant time-delay, the UT-based CISM indirect update algorithm is proposed. Limited by the performance of the 2-D turntable and the measured-correlation of the two sensor systems, the measurement lag of the laser range finder system is inevitable. Combining the UT algorithm and the complementary advantages of multi-sensor information, the time-delay information of the laser range sensor is updated by using the measured information of binocular vision sensor at the time-delay moment. After the time-delay information is arriving at the fusion center, UT-A1 is combined to solve the problem of CISM. Simulations show that the proposed algorithm improves the real-time performance of the system effectively.(5) Based on the researches of the fusion algorithm with estimate correlation, on the basis of the fourth part, the no-resent mode hybrid federated kalman fusion algorithm with independent information sharing is proposed. Due to the measured-correlation of the two sensor systems, it is difficult to given the cross-covariance in the process of information fusion. The CI-Kalman fusion algorithm with diagonal matrix weighting is proposed to solve this problem, but the estimation accuracy is decreased. Then, based on the federated kalman filter, the information sharing is redistributed and the new improved information fusion algorithm is proposed which is combining the no-rest model and UT-CISM direct update algorithm. The simulations show that the algorithm has high accuracy and good applicability to the system.(6) We designed and established the experiment system of heterogeneous target tracking system. The hardware equipments were selected and the control software was completed. Finally the corresponding experiments were carried out for each algorithm proposed above. The experimental results illustrated the feasibility of the proposed system and the accuracy and effectiveness of the proposed algorithm.Through the research on the various elements of the target tracking method, the problems such as the outlier-elimination, the motion state modeling and the measured-correlation are solved, and the purpose of improving the multi-sensor tracking system precision is achieved. The researches of this paper have important engineering significance of the oil inlet tracking in autonomous air refueling and have some academic significance to Improve of target tracking and information fusion.