The Research Of Multi-source Data Association And Fusion Algorithm

Multi-source data fusion which can be called multi-sensor information fusion is a new technology for processing from multi-sensor data, including detection, correlation, combination, estimation, it is of great significance to improve the accuracy of status and identity estimation, compared with the single sensor, the accuracy and reliability of data is higher, the technology is widely applied in military and civil fields.At present, there are still many key problems in the research of multi-source data fusion technology. This paper discusses and studies the problems in two aspects. On the one hand is the data association problem: as one of the key problems of multi-source data fusion, under the multi-target environment, the current data association algorithm is faced with the problem of large amount of computation, poor real-time performance of tracking and so on. To reduce the computational complexity of the algorithm has a certain practical significance to improve the real-time performance of the algorithm and engineering application. On the other hand is the data fusion problem: In the process of detecting data, the multi sensors are affected by many factors, which makes the data to be fused, The uncertainty includes not only the measurement uncertainty and noise, also but includes the fuzziness and inconsistencies in the actual environment, research on how to improve the accuracy and reliability of the fusion result is of certain value.In this paper, data association and data fusion is used as the research object, the main research content includes the following aspects:(1) This paper introduces the theory of multi-source data fusion, the model structure and the three basic levels of fusion, summarizes the problem the multi-source data fusion technology, and expounds the existing multi-source data association and data fusion technology.(2) Based on the theory of data association, three kinds of algorithms are compared and analyzed, which mainly focus on the nearest neighbor data association, the probability data association, and the joint probability data association, the advantages and disadvantages and application scope of the different simulation analysis are verified by different simulation analysis.(3) Aiming at the problem of the high computational burden and poor real-time performance of tracking in the JPDA algorithm, which leading to a limit in engineering application, considering the deficiencies of PDA and JPDA, this paper proposes an improved algorithm which based on the fast data association algorithm that based on the public measurement in the intersection range gate of multi-target for processing. In the improved algorithm, the problem of high computational burden and poor real-time performance of tracking in the JPDA algorithm is solved to a certain extent by the thought of calculating the probability which decided by the distance between the public measurement and each target. In this way, the improved PDA algorithm can be used in the multi-target tracking environment, and has certain significance for the practical engineering applications.(4) From the perspective of improving the data fusion results precision, due to the multi-sensor measurement of many times on some target characteristic parameters with unknown prior knowledge, a new and reliable fusion method based on connection degree of set pair analysis is proposed. This method utilizes the advantages of the characteristic function of the set pair analysis theory, and defines the relation matrix through the mining of measured data of the opposite, the same and the difference to adjust the relation between data, and the connection degree matrix's dimension is augmented to measure the mutual integrated contact degree of sensors at different moments, then according to the ratio of signal to noise which has existed, the measurement data is reasonably assigned weights in the process of integration, and then the fusion formula of multi-sensors data is obtained. The application examples and simulation results prove that the proposed method has better accuracy and reliability.