Multi-target tracking method based on graph network

Abstract

The invention provides a multi-target tracking method based on a graph network, and the graph network comprises a feature extraction network module and a graph convolution matching module. The methodspecifically comprises the following steps: S1, selecting two frames of images in a video, and inputting the two frames of images into the feature extraction network module; S2, performing feature extraction on the targets through a feature extraction network module, and acquiring target feature vector sets FM and FN of the two frames of images, wherein M and N represent the number of the targetsdetected in the two frames of images respectively; S3, calculating the similarity between the target feature vectors based on the target feature vector sets FM and FN, and constructing a bipartite graph; S4, matching the bipartite graph through the graph convolution matching module, and performing back propagation of the graph network by using a loss function to obtain an optimal matching matrix.According to the invention, a convolutional network is utilized to perform feature extraction on the targets, the proposed loss function solves the problem of uncertain target number, and the accuracyof multi-target tracking is greatly improved.

Description

technical field [0001] The invention relates to the technical field of computer vision tracking, in particular to a graph network-based multi-target tracking method. Background technique [0002] The target tracking technology observes the position information of the moving target, connects the targets in series according to the time sequence, analyzes the behavior of the moving target by using the obtained trajectory, and predicts the movement of the target at the next moment, which can be used to predict the future movement trend of the target or the target Behavioral status has very important value in many application scenarios. For example, in machine navigation, automatic driving and video surveillance systems, online multi-target tracking technology plays an important role. [0003] For the patent application CN201910429444.X, an online multi-target tracking method based on deep learning and data association is disclosed, which includes the following steps: 1. Input t...

AI Technical Summary

Problems solved by technology

However, directly calculating the similarity between targets and using the Hungarian algorithm for trajectory matching cannot correct problems such as false detections and missed detections. In the scene where targets are occluded or similar targets are intersected, identities are easily exchanged, resulting in an increase in the number of IDSw

Images