Modified Particle Filter Based on Probabilistic Hypothesis Density Phd

Abstract

The invention provides a particle PHD filter for tracking multiple targets when the intensity of the new target is unknown. The filter first uses threshold technology to classify the measured values, and separates the measured values ​​of the surviving targets from the measurement set for the PHD update of the surviving targets, while the remainder set is discarded; and then distinguishes the surviving targets and new target, and determine the new target generation point based on the measurement value when the new target intensity is unknown, and decompose the target PHD into new target PHD and surviving target PHD; finally, use two different particle sets to separate the surviving target PHD and the new target PHD Particle approximation is carried out, and after prediction, update and resampling, the particle realization of the proposed PHD filter is carried out. The filter overcomes the defect that the standard particle PHD needs to a priori the strength of the newborn target, and because the clutter information is eliminated, the real-time performance of the filter can be improved while improving the performance.

Description

technical field [0001] The invention relates to the field of multi-target tracking in data fusion processing technology, and is suitable for multi-target tracking scenarios when the number of targets changes in real time, measurement information has great uncertainty, and the strength of new-born targets is unknown. Background technique [0002] In the actual target tracking scene, due to the derivation, disappearance and appearance of new targets, the number of targets changes in real time, and the measurement information also has great uncertainty (targets, clutter or false alarms, etc.), for Multiple object tracking poses enormous difficulties. Traditional multi-target tracking methods usually assume that the number of targets is known or unknown. Through data association, the multi-target tracking problem is transformed into a single-target filtering tracking problem. However, when the targets are dense or there are many false alarms, data association will bring about a ...

AI Technical Summary

Problems solved by technology

The standard PHD filter has the following two problems: one is the implementation of the standard filtering algorithm, especially the real-time implementation problem, and the other is that the standard PHD filter assumes that the strength of the new target is known, which is contrary to the actual background

[0004] In view of the above two problems, it is necessary to study the PHD particle filter when the strength of the newborn target is unknown, so as to solve the multi-target tracking problem in the actual background

Images