A Data-Driven Consistent Object Tracking Method

Abstract

A data-driven consistent target tracking method, organizes and defines the data sequence of each node, initializes system parameters, and understands the distribution of node locations; uses time functions to match node data, and converts node distribution into the form of a general network topology map; Perform optimization calculations separately to obtain the fitting parameters of each function, and then select time t to substitute into the fitting function to obtain the estimated data of each node, and obtain the corresponding Laplacian matrix from the topological map; calculate from the Laplacian matrix The weight matrix is ​​substituted into the iterative algorithm to make the data of each node consistent as soon as possible. Aiming at the moving target without model, the present invention uses a time function fitting method to obtain the state estimation through multi-sensor data and distribution position, and then calculates according to the node position so that the data of each node can quickly reach the same level, effectively utilizing the sensor Interactive technology improves estimation accuracy.

Description

technical field [0001] The invention belongs to the field of target tracking, and relates to estimating the target trajectory by using the measured values ​​of multiple sensor nodes, and then making the estimation of each node consistent by means of distributed calculation. Background technique [0002] In recent years, with the rapid development of sensors and communication technologies, research results in the field of target tracking are changing rapidly, and are widely used in practical applications such as robot collaborative work, Internet of Vehicles, and security monitoring. Generally, most existing target tracking algorithms can be divided into two categories. One is based on "model-driven". First, the single-model or multi-model method is used to match the target's motion state, and then the calculation is performed according to the matching model to obtain the target's motion Trajectories; the other is "data-driven" that relies on observation data, using maximum l...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the problems of difficult acquisition of models and high computational complexity of consistent processing in the current tracking problem, the present invention proposes a method of using time function curve to fit the observation trajectory, and then make each sensor achieve data consistency as soon as possible according to the sensor position distribution. The processing method improves the tracking accuracy and ensures the reliability of subsequent operations

Images