Distributed fusion method of label random set filter based on matching of label spaces

Abstract

The invention discloses a distributed method of a label random set filter based on matching of label spaces. Label random set distribution, from different sensors, corresponding to set marginal probability density distribution of different object labels is calculated; an arrangement distribution problem related to the matching relation of different label spaces is constructed by designing a corresponding cost function and distribution matrix and the like; the arrangement distribution problem is solved to obtain an optimal distribution matrix or mapping relation, and matching of the different label spaces is realized; and under matching of the label spaces, multi-sensor posterior probability fusion is carried out on the basis of a generalized covariance cross-information criterion. According to the method, a parallel fusion structure is designed for probability density functions of the object labels, the fusion algorithm is characterized by high efficiency, high operation efficiency, lower engineering realization complexity and the like, a mixed Gaussian model is used to reduce the engineering realization complexity, and the problem that a fusion result collapses due to mismatching of the label spaces is solved.

Description

technical field [0001] The invention belongs to the field of multi-sensor fusion, in particular to the technical field of multi-target tracking and distributed multi-sensor fusion under random set theory. Background technique [0002] Compared with centralized multi-sensor fusion technology, distributed multi-sensor fusion technology has the advantages of low communication cost and strong system fault tolerance, so it has received more and more attention. widely. Many scholars have done a lot of research on the problems and difficulties in multi-sensor fusion, and derived a variety of distributed sensor fusion algorithms: MHT, JPDA, CPJFA and other algorithms. These algorithms are often based on the assumption that sensors are independent of each other, but in most practical scenarios, this assumption is not valid. When different sensors observe the same target, there are common process noise and measurement noise among the sensors, and these common information lead to unk...

AI Technical Summary

Problems solved by technology

These algorithms are often based on the assumption that the sensors are independent of each other, but in most practical scenarios, this assumption is not valid

When different sensors observe the same target, there are common process noise and measurement noise between the sensors. These public information lead to unknown level correlations between the measurements from different sensors.

[0003] In the field of traditional multi-sensor tracking, although scholars have done relevant research based on the assumption of correlation between sensors, there are some problems in the traditional multi-sensor multi-target fusion algorithm: 1) There are a large number of data association algorithms: Data association and data association of targets between sensors; 2) Due to the correlation between sensors, a large number of calculations are required to calculate the correlation when performing target fusion

However, there is a major challenge for the distributed fusion of label random set probability density functions: the label space mismatch phenomenon of label set probability density functions from different sensors, which has been proved by experiments to occur frequently in actual scenarios and is a common phenomenon

Images