Abstract：The detection and tracking of maritime targets are key tasks in maritime security and military applications. To address the low measurement accuracy of target distance in complex maritime environments, affected by factors such as sea clutter, an improved multi-sensor data fusion algorithm is proposed. The algorithm utilizes the joint detection results of shipborne radar and land-based radar on the sea surface. Firstly, the data from multiple sources are transformed into a common coordinate system. Then, the Robust Z-score method is employed for preprocessing the longitudinal data to eliminate abnormal data. Subsequently, by redefining the confidence distance measure, sensor results with higher confidence are used to replace the discarded data, and an adaptive weighted fusion of the results is performed. Furthermore, to further improve the data accuracy, a segmented fusion mechanism is introduced. The improved sensor data fusion algorithm is cascaded with a stepwise adaptive weighted fusion algorithm. By measuring the similarity of the fusion results in each segment and setting a confidence threshold, the final fusion result is determined. Simulation experiments confirm the effectiveness and accuracy of the algorithm.

Abstract：Due to the low tracking accuracy of existing simple transition probability matrices and the long tracking time of complex transition probability matrices, it is difficult to meet the requirements of maneuvering target tracking in three-dimensional space. Aiming at the design problem of transition probability matrix, starting from mechanism analysis, a model transition probability matrix design method based on membership function is proposed, and the three-dimensional interactive multi model algorithm is improved and perfected. The simulation results show that the method of modifying the transition probability matrix based on the membership function effectively improves the tracking accuracy of three-dimensional maneuvering targets.

Abstract：To address the issue of poor target state estimation accuracy under conditions where sensors have system bias and noise is non-Gaussian, a target tracking method based on the maximum correntropy Kalman filter (MCKF) under biased measurements is proposed. This approach introduces a differential mechanism that constructs differential measurement equations from the biased measurements of the target at adjacent time points, effectively mitigating the effects of system bias. Subsequently, the maximum correntropy criterion (MCC) is employed to quantify the higher-order moment information of the estimation error with differential measurements serving as a priori conditions. This leads to the derivation of the filtering iterative equations for the algorithm under biased measurements. Simulation results demonstrate that, when system observations are affected by sensor system bias and non-Gaussian noise, the proposed method outperforms existing approaches in terms of tracking performance.

Abstract：The lag of maneuvering target tracking results on the time axis is a major difficulty in the field of maneuvering target tracking. There are many cases of time delay, which are especially obvious in the early stage of tracking and the time period when a large maneuver occurs, and there is often a peak of error because of this, if there is a way to suppress or eliminate the time delay phenomenon, the tracking effect will be significantly improved. Starting from the results and phenomena of simulation experiments, combined with Kalman filter theory, interactive multi-model algorithm and modern neural network model, this paper will deeply analyze the time delay problem, and obtain different causes of time delay according to the changes of each stage of tracking and give corresponding solutions, so as to provide reference for improving the tracking effect of maneuvering targets in the future.