Adjacent multi-target tracking method based on Gaussian mixture probability hypothesis density

Abstract

The invention discloses an adjacent multi-target tracking method based on Gaussian mixture probability hypothesis density. The method comprises the following steps: adding a label and a historical state matrix as auxiliary parameters to construct a new standard description set of a target; initializing a target probability hypothesis density, a target label set and a target historical state matrixset; calculating target prediction probability hypothesis density, a target prediction label set and a target prediction historical state matrix set according to the probability hypothesis density, the label set and the historical state matrix set of the new target and the survival target; calculating a target posterior probability hypothesis density, a target posterior label set and a target posterior historical state matrix set based on the measurement set, and reallocating the weight of each Gaussian component in the target posterior probability hypothesis density; transforming the Gaussian component set and the parameter set of the target, and reducing the transformed Gaussian component set; estimating the state and number of targets; and if tracking is performed at a single moment, ending tracking; and if a plurality of moments are tracked, iterating all moments. The method has good tracking performance and robustness.

Description

Technical field [0001] The invention belongs to the technical field of intelligent information processing, and specifically relates to a method for tracking multiple targets in close proximity based on Gaussian mixture probability hypothesis density. Background technique [0002] In recent years, the Probability Hypothesis Density (PHD) filter based on the finite set statistical theory has greatly reduced the computational complexity because it does not require a complex data association process, and has attracted widespread attention from scholars in the field of multi-target tracking. [0003] PHD filter is an approximation method of multi-objective Bayesian filter. What it transmits at each moment is not the complete posterior density of the target, but the probability hypothesis density of the target (the first-order statistics of the complete posterior density of the target Moment), the target state and number are obtained from the target probability hypothesis density. Howev...

AI Technical Summary

Problems solved by technology

[0018] Aiming at the problem of low target state and number estimation accuracy of the multi-target tracking method based on PHD filtering in the parallel moving target scene, the present invention proposes a method based on Gaussian mixture Probability Hypotheses Density Probable-Multi-Target Tracking Method, using Gaussian Mixture Probability Hypothesis Density (MCST-GM-PHD) to solve the problem of parallel moving target tracking in dense clutter and low detection probability tracking environment

Images