A Modeling Method Tracking Method and Device

Abstract

Embodiments of the present invention provide a graph neural network target tracking method and device. The method includes: determining a target tracking graph structure data generation method, converting and generating early warning detection data into graph structure data; constructing a target tracking graph neural network, including graph representation There are two parts: processing the generated graph and generating the judgment results of nodes and edges; setting the target tracking graph neural network loss function for network training optimization; collecting early warning detection data in different environments, and establishing target tracking Graph neural network training data set; using gradient backpropagation algorithm to optimize the parameters of the graph neural network by minimizing the set loss function; deploying the trained target tracking graph neural network, The probability of the edge is judged, and the real target measurement and the real target track are obtained. The embodiment of the present invention provides a unified and efficient target tracking method, which can realize stable, fast and accurate tracking of the target.

Description

technical field [0001] The invention relates to target tracking technology, more particularly, the invention relates to a graph neural network target tracking method and device, which are suitable for single-target or multi-target tracking problems of radar, optical, infrared and other early warning detection equipment. Background technique [0002] Target detection is a technical activity that uses radar, optical, infrared and other early warning detection equipment to discover, locate and track targets. The results can be used as input and basis for further high-level processing such as attribute identification, threat assessment and operational decision-making. Target tracking is an important link in the target detection process. Its purpose is to connect the detection information from the same target at different times, and through filtering estimation, form time-series target status information with the same identity, that is, the target track. Real-time, continuous and...

AI Technical Summary

Problems solved by technology

However, when faced with multi-type complex tracking scenarios such as weak, dense, and formations, or mixed tracking scenarios composed of multiple single-type scenarios, the tracking effect of existing methods is often not good due to their simple models, limited capabilities, and insufficient generalization. Stable, sometimes good and sometimes bad, it is difficult to achieve stable and good tracking effects in various situations and in the whole scene, it can only solve the problem of target tracking in specific situations and local scenes that match the model

At the same time, the existing target tracking methods still have the problem that the parameters cannot be determined, and manual modification and debugging are required, which will consume a lot of time and energy, and the tracking effect after debugging is difficult to achieve optimal

It can be seen that the existing radar target tracking methods have the problems of simple model, low complexity, poor applicability, and lack of learning ability, so it is difficult to solve the target tracking problem as a whole

In addition, for optical and infrared target tracking, since optics and infrared can obtain more target appearance feature information, through target appearance feature matching, effective tracking of targets can be achieved in some scenes, but there are also problems of similar target appearance and mutual Various challenges such as occlusion and dramatic changes in target appearance

Images