Research On The Fine Grained Evaluation Index Of Multi Sensor Information Fusion Algorithm

At present, information fusion technology in China is booming, and widely used in various fields. Accordingly, disturbing technology to the interference of fusion is also in rapid development, at the same time, the fusion effect needs more and more high requirements. Since the fusion algorithm has been relatively mature, we choose the point of algorithm management to improve the fusion system. We divide the whole fusion process into four coarse granularity steps:state estimation, filtering threshold, data association and track fusion. For every step, there are many algorithms can be selected. Although these algorithms achieve the same functionality, but their respective characteristics and application environment are different. How to select the parameters and the optimal algorithm to make the final fusion effect to be optimal, according to the different environment data sources, has been one of the research hotspots.This paper mainly focuses on the extraction of fine-grained algorithm unit module and the construction of sensitive index system of multi granularity. Mainly include the following:(1) The original information fusion coarse-grained evaluation index can be reflected in the state estimation, filtering threshold, date association and track fusion. So this paper focuses on the four algorithms and the unit module of them. Based on the unit module algorithm of state estimation and date association, we study the track fusion algorithm deeply, then we can analysis the performance and difference between sub algorithm on the fine-grained level.(2)In the original information fusion of coarse granularity index:the data source evaluation index(20index), tack quality index(27index), track fusion performance evaluation index(18index). This paper put forward the fine grained index to the total return wave number of "echo number index", the innovation vector norm of filtering threshold; the fine grained index to the effective number of echoes of "echo number index", the metrics based on uncertainty of data association; the fine grained index to the mean error distance of the track precision index, the GOSPA index of track fusion.(3)Based on the unit module algorithm of the track fusion, we improved the metrics global optimal sub-pattern assignment (GOSPA) distance measure. By experiment verification, due to considering the uncertainty, the metric HGOSPA can make the correct evaluation of different track fusion algorithms without the true value.