Method for tracking multiple targets under unknown measurement noise distribution

Abstract

The invention relates to the technical field of target tracking, in particular to a method for tracking multiple targets under unknown measurement noise distribution. Through the advantages and disadvantages of a risk value measurement particle, the particle is evaluated through a risk evaluation function and weight updating is performed. The entire process is independent of the measurement noise distribution and multiple targets can be tracked stably under unknown complex measurement noise distribution. The method has higher robustness and stability compared with those of the ordinary particle probability hypothesis density filtering technology.

Description

technical field [0001] The invention relates to the technical field of target tracking, in particular to a multi-target tracking method under unknown measurement noise distribution. Background technique [0002] The purpose of target tracking is to use the tracking filter algorithm to estimate the moving state of the moving target such as position, velocity and acceleration in real time. Compared with single-target tracking, multi-target tracking becomes more complex due to the time-varying number of targets, the uncertainty of the correspondence between observations and targets, and observation clutter. [0003] There are two main categories of existing multi-target tracking technologies, one is the traditional multi-target tracking technology, and the other is the multi-target tracking technology based on Random Finite Set. Traditional multi-target tracking technology needs to determine the correspondence between observations and targets through data association, so it is...

AI Technical Summary

Problems solved by technology

Traditional multi-target tracking technology needs to determine the correspondence between observations and targets through data association, so it is also called association-based multi-target tracking technology. This type of technology has some shortcomings: first, tracking The accuracy of the data depends largely on the accuracy of the data association. At the same time, the data association takes a lot of time and affects the speed of tracking.

[0005] However, when the particle probability hypothesis density filtering technology implemented according to the above four steps is used for tracking filtering, it needs to rely on the probability distribution of the observation noise to calculate the likelihood function to update the particle weights. However, in the actual tracking process, the measurement noise is often complicated. It is difficult to accurately obtain the probability distribution of the measurement noise. If the particle weight is updated with an inaccurate noise distribution that does not match the actual measurement noise, the tracking performance of the algorithm will be reduced or even diverge.

On the other hand, the filter update needs to calculate the likelihood function in the multi-dimensional observation space, which will consume a lot of time and reduce the tracking speed

Images