CN-121980129-A - Distributed target fusion tracking method and device based on space-time registration

CN121980129ACN 121980129 ACN121980129 ACN 121980129ACN-121980129-A

Abstract

The invention provides a distributed target fusion tracking method and device based on space-time registration, and relates to the technical field of sensors. The distributed target fusion tracking method based on space-time registration comprises the steps of selecting the mean value of a Gaussian component with the largest weight from a plurality of Gaussian components of a posterior probability density expression of space-time registration parameters as an estimated value of the space-time registration parameters, aligning second posterior probability density with first posterior probability density in time and space according to the estimated value of the space-time registration parameters by utilizing a space-time registration formula of a uniform motion model to obtain calibrated second posterior probability density, and fusing the first posterior probability density with the calibrated second posterior probability density based on a generalized covariance cross fusion criterion to obtain target existence probability and target space density of a tracking target. The invention can realize the registration of the multi-node sensor data with high precision and high reliability.

Inventors

LI GUCHONG
ZHANG YUCHEN
LI TIANCHENG
WANG XUEQIAN
LI GANG
LI YAOWEN

Assignees

西北工业大学
清华大学
清华大学深圳国际研究生院

Dates

Publication Date: 20260505
Application Date: 20251229

Claims (10)

1. A distributed target fusion tracking method based on space-time registration is characterized by comprising the following steps: based on a generalized covariance cross fusion criterion, constructing a posterior probability density expression of a space-time registration parameter by using a uniform motion model and a Bernoulli filter, wherein the posterior probability density expression of the space-time registration parameter is in a Gaussian mixed form, and the space-time registration parameter is used for realizing alignment of a first posterior probability density and a second posterior probability density in a time dimension and a space dimension, wherein the first posterior probability density is obtained by carrying out Bernoulli filtering on data acquired by a reference sensor, and the second posterior probability density is obtained by carrying out Bernoulli filtering on data acquired by a sensor to be registered; selecting the mean value of the Gaussian component with the largest weight from a plurality of Gaussian components of the posterior probability density expression of the space-time registration parameter as the estimated value of the space-time registration parameter; Aligning the second posterior probability density with the first posterior probability density in time and space according to the estimated value of the space-time registration parameter by using a space-time registration formula of the uniform motion model to obtain a calibrated second posterior probability density; Based on a generalized covariance cross fusion criterion, fusing the first posterior probability density and the calibrated second posterior probability density to obtain the target existence probability and the target space density of the tracking target; And determining target state estimation of the tracking target according to the decision threshold, the target existence probability and the target space density of the tracking target.
2. The distributed object fusion tracking method based on space-time registration according to claim 1, wherein the constructing a posterior probability density expression of space-time registration parameters based on generalized covariance cross fusion criteria by using a uniform motion model and a bernoulli filter comprises: Constructing a posterior probability density expression of a time registration parameter and a spatial registration parameter based on a generalized covariance cross fusion criterion, wherein the time registration parameter is used for calibrating the time of the second posterior probability density based on the reference time corresponding to the first posterior probability density, and the spatial registration parameter is used for calibrating the spatial coordinate of the second posterior probability density based on the reference spatial coordinate corresponding to the first posterior probability density; Coupling the time registration parameter and the space registration parameter in the posterior probability density expression of the time registration parameter and the space registration parameter into the space-time registration parameter by utilizing a uniform motion model to obtain the posterior probability density expression of the space-time registration parameter; And transforming the posterior probability density expression of the space-time registration parameter into a Gaussian mixture form by using a Bernoulli filter and a Gaussian mixture approximation technology to obtain the posterior probability density expression of the space-time registration parameter in the Gaussian mixture form.
3. The distributed object fusion tracking method based on space-time registration according to claim 2, wherein the constructing a posterior probability density expression of a temporal registration parameter and a spatial registration parameter based on a generalized covariance cross fusion criterion comprises: Constructing a prior probability density function between the first posterior probability density and the second posterior probability density with respect to a temporal registration parameter and a spatial registration parameter; Constructing a joint likelihood function between the second posterior probability density and the second posterior probability density with respect to a temporal registration parameter and a spatial registration parameter; And constructing posterior probability density expressions of the time registration parameters and the space registration parameters according to the prior probability density function and the joint likelihood function.
4. The method for distributed object fusion tracking based on space-time registration according to claim 3, wherein the coupling the temporal registration parameter and the spatial registration parameter in the posterior probability density expression of the temporal registration parameter and the spatial registration parameter to the space-time registration parameter by using a uniform motion model, to obtain the posterior probability density expression of the space-time registration parameter, comprises: Coupling the time registration parameter and the space registration parameter in the prior probability density function of the time registration parameter and the space registration parameter into a space-time registration parameter by utilizing a uniform motion model to obtain the prior probability density function of the space-time registration parameter; coupling the time registration parameter and the space registration parameter in the joint likelihood function of the time registration parameter and the space registration parameter into a space-time registration parameter by utilizing a uniform motion model to obtain the joint likelihood function of the space-time registration parameter; And constructing a posterior probability density expression of the space-time registration parameter according to the prior probability density function of the space-time registration parameter and the joint likelihood function of the space-time registration parameter.
5. The method for distributed object fusion tracking based on space-time registration according to claim 4, wherein the transforming the posterior probability density expression of the space-time registration parameter into a gaussian mixture form by using a bernoulli filter and a gaussian mixture approximation technique to obtain the posterior probability density expression of the space-time registration parameter in the gaussian mixture form comprises: using Gaussian mixture technology, expressing the priori of the target state of the tracking target as Gaussian mixture, substituting the Gaussian mixture into the joint likelihood function of the space-time registration parameter to obtain the joint likelihood function of the space-time registration parameter of the Gaussian mixture; And obtaining a posterior probability density expression of the space-time registration parameters of the Gaussian mixture form according to the joint likelihood function and the prior probability density function of the space-time registration parameters of the Gaussian mixture form.
6. The distributed object fusion tracking method based on space-time registration according to claim 5, wherein the posterior probability density expression of the space-time registration parameter is as follows: ; Wherein, the ; ; ; ; Where L is the number of Gaussian components, i and j are the number of sensors, The weight coefficient related to the existence probability after the first posterior probability density and the second posterior probability density are fused; a weight coefficient for the first gaussian component; As a gaussian distribution function, wherein Is a space-time registration parameter that is used to determine, For the mean vector of the first gaussian component, Covariance matrix of the first Gaussian component; A is the number of the a-th Gaussian component in the sensor i, and b is the number of the b-th Gaussian component in the sensor j; Is the number of gaussian components in sensor i; is the number of gaussian components in sensor j; is a fusion weight; is the mean value of the a-th Gaussian component in the sensor i; the mean value of the b-th Gaussian component in the sensor j; the mean vector is calculated by the first Gaussian component in the prior probability density of the space-time registration parameter, the a-th Gaussian component in the sensor i and the b-th Gaussian component in the sensor j; the covariance matrix is obtained by calculating the first Gaussian component in the prior probability density of the space-time registration parameter, the a-th Gaussian component in the sensor i and the b-th Gaussian component in the sensor j; is the mean value vector of the first Gaussian component in the Gaussian mixture representation of the prior density of the space-time registration parameter, wherein I is an identity matrix; the covariance matrix of the first Gaussian component in the Gaussian mixture representation of the space-time registration parameter prior density is obtained; is the covariance matrix associated with the a-th gaussian component in sensor i and the b-th gaussian component in sensor j.
7. A distributed target fusion tracking device based on space-time registration, comprising: The construction module is used for constructing a posterior probability density expression of a space-time registration parameter by utilizing a uniform motion model and a Bernoulli filter based on a generalized covariance cross fusion criterion, wherein the posterior probability density expression of the space-time registration parameter is in a Gaussian mixture form, the space-time registration parameter is used for realizing alignment of a first posterior probability density and a second posterior probability density in a time dimension and a space dimension, the first posterior probability density is obtained by carrying out Bernoulli filtering on data acquired by a reference sensor, and the second posterior probability density is obtained by carrying out Bernoulli filtering on the data acquired by a sensor to be registered; the selection module is used for selecting the mean value of the Gaussian component with the largest weight from a plurality of Gaussian components of the posterior probability density expression of the space-time registration parameter as the estimated value of the space-time registration parameter; The calibration module is used for aligning the second posterior probability density with the first posterior probability density in time and space according to the estimated value of the space-time registration parameter by using a space-time registration formula of the uniform motion model to obtain a calibrated second posterior probability density; the fusion module is used for fusing the first posterior probability density and the calibrated second posterior probability density based on a generalized covariance cross fusion criterion to obtain the target existence probability and the target space density of the tracking target; and the determining module is used for determining the target state estimation of the tracking target according to the judgment threshold value, the target existence probability and the target space density of the tracking target.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that the processor implements a distributed object fusion tracking method based on space-time registration as claimed in any one of claims 1 to 6 when executing the computer program.
9. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements the distributed target fusion tracking method based on space-time registration according to any of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements a distributed object fusion tracking method based on space-time registration as claimed in any one of claims 1 to 6.

Description

Distributed target fusion tracking method and device based on space-time registration Technical Field The invention relates to the technical field of sensors, in particular to a distributed target fusion tracking method and device based on space-time registration. Background The target tracking technology is a core support technology in the fields of intelligent transportation, unmanned systems, security monitoring and the like. In intelligent traffic, the intelligent traffic system can track vehicles and pedestrians in real time, optimize traffic flow and prevent accidents, an unmanned system runs autonomously and safely by means of the intelligent traffic system, security monitoring relies on the intelligent traffic system to discover anomalies in time, and public safety is guaranteed. Along with the expansion of application scenes, the scale of the sensor network is continuously enlarged and the isomerism is enhanced. The sensors with different types and performances work cooperatively, and the system needs to fuse multiple sections of sensor data so as to realize the robust tracking of the dynamic target and improve the accuracy and reliability of the tracking. However, existing approaches have technical bottlenecks in time registration and sensor offset registration (spatial registration). The reasons for the space-time misregistration problem mainly include that firstly, the sampling rate of sensor nodes is various, or communication and processing have delay to cause time deviation, and secondly, the position of a sensor is unknown to cause space deviation. If the errors are not estimated and compensated timely and accurately, the fusion precision of the multi-sensor system can be reduced, and the reliability of target state estimation is affected. The related technology is limited to registration distributed fusion tracking with single dimension in time or space, has obvious limitation in actual complex scene application, and is difficult to meet the target tracking requirement of high precision and high reliability. Therefore, how to realize the registration of the multi-node sensor data with high accuracy and high reliability is a technical problem to be solved. Disclosure of Invention The invention provides a distributed target fusion tracking method and device based on space-time registration, which are used for solving the defects in the prior art and realizing registration of multi-node sensor data with high precision and high reliability. The invention provides a distributed target fusion tracking method based on space-time registration, which comprises the following steps of. The method comprises the steps of constructing a posterior probability density expression of a space-time registration parameter based on a generalized covariance cross fusion criterion by using a uniform motion model and a Bernoulli filter, wherein the posterior probability density expression of the space-time registration parameter is in a Gaussian mixture form, the space-time registration parameter is used for realizing alignment of a first posterior probability density and a second posterior probability density in a time dimension and a space dimension, the first posterior probability density is obtained by carrying out Bernoulli filtering on data acquired by a reference sensor, the second posterior probability density is obtained by carrying out Bernoulli filtering on the data acquired by the reference sensor, a mean value of a Gaussian component with the largest weight is selected from a plurality of Gaussian components of the posterior probability density expression of the space-time registration parameter to serve as an estimated value of the space-time registration parameter, the second posterior probability density is aligned with the first posterior probability density in time and space according to the estimated value of the space-time registration formula of the space-time registration parameter, and the target tracking probability density is obtained by using the generalized covariance cross probability, and the target tracking probability density is determined based on the generalized probability threshold, and the target tracking probability density is obtained by using the generalized probability. According to the distributed target fusion tracking method based on space-time registration, the posterior probability density expression of space-time registration parameters is constructed by utilizing a uniform motion model and a Bernoulli filter based on generalized covariance cross fusion criteria, and the method comprises the following steps: The method comprises the steps of constructing a posterior probability density expression of a time registration parameter and a space registration parameter based on a generalized covariance cross fusion criterion, wherein the time registration parameter is used for calibrating time of the second posterior probability density based on reference time corresponding to