Heterogeneous sensor information fusion method

Abstract

The invention discloses a heterogeneous sensor information fusion method. The method comprises a local estimation part and a fusion estimation part. Each local sensor calculates pseudo measurement toform a local measurement spectrum, and obtains a generalized covariance intersection of local estimation. In the fusion estimation part, based on the Sklar theorem, a Copula function representing thedependency relationship between modes is used for estimating the correlation relationship between random variables of different sensors; optimizing the correlation coefficient of the Copula function according to the minimum resolution information criterion by comparing the Kullback-Leibler divergence between the fusion estimation and the local estimation; performing importance sampling by using the generalized covariance intersection to improve the calculation efficiency, and constructing Gaussian approximation of fusion density by using kernel density estimation; and recursively updating thefusion estimation in a layered manner to realize the fusion of any number of heterogeneous sensors. According to the invention, the requirements on the communication rate and correlation calculation of the local sensor are reduced.

Description

technical field [0001] The invention relates to the technical field of data fusion, in particular to a heterogeneous sensor information fusion method. Background technique [0002] Distributed state estimation of dynamic systems is an important technology in the field of digital signal processing, and has been widely used in cooperative target tracking, distributed formation of space telescopes, and remote environmental monitoring. At present, the rapid development of wireless sensor hardware with advanced sensing, computing and communication capabilities has greatly promoted the development of data fusion technology in distributed estimation systems. The sensor fusion approach provides a greater sensing range than a single sensor by expanding the geographic range and improving filtering accuracy. [0003] "Comparison of Two-sensor Tracking Methods Based on State Vector Fusion and Measurement Fusion", published in "IEEE Transactions on Aerospace and Electronic Systems" (J R...

AI Technical Summary

Problems solved by technology

However, existing sensor fusion methods rely on known correlations, so they cannot optimally combine correlated data from different sources, and these suboptimal fusion algorithms are conservative in their results and have low estimation accuracy

There is currently no general modeling approach to construct joint probability densities between the intermodal dependency structures and observations from statistically correlated sensors

Images