Multi-target tracking method and face recognition method

Abstract

The invention provides a multi-target tracking method and a face recognition method. The multi-target tracking method comprises the steps of obtaining a to-be-processed video image, wherein the to-be-processed video image comprises a plurality of frame images; respectively carrying out target identification processing on each frame of image to obtain position related information and feature information of each target in each frame of image, wherein at least one frame of image comprises a plurality of targets; analyzing and processing each target in each frame of image to judge whether each target belongs to a specific target or not; and when the target belongs to the specific target, carrying out re-identification processing according to the feature information or according to the positionrelated information and the feature information at the same time. According to the invention, high accuracy of multi-target tracking can be realized.

Description

technical field [0001] The present invention relates to the technical field of computer vision, in particular to a multi-target tracking method, a face recognition method, a computer-readable storage medium and computer equipment. Background technique [0002] As an important part of computer vision applications, video tracking technology is widely used in environmental monitoring, human-computer interaction, and robot navigation. According to different feature values, the same moving object between consecutive frames in the image sequence is associated to obtain the motion parameters of the object in each frame image and the corresponding relationship between the moving objects in adjacent frame images, and further obtain the moving object complete motion track. [0003] When solving the multi-target tracking problem, in order to save computational costs, it is usually divided into two steps: [0004] The first step is to perform target recognition processing on each fram...

AI Technical Summary

Problems solved by technology

[0008] (1) Only use the position-related information X to perform one-step ReID. At this time, the speed is faster, but the matching accuracy is low, and the detection requirements are high;

[0009] (2) Only use the feature information F to perform one-step ReID. At this time, the matching accuracy is high, but the speed is slow, and it is especially difficult to achieve real-time tracking of multiple targets;

[0010] (3) Use location-related information X and feature information F at the same time to perform one-step ReID (for example, by mixing two parameters with a certain weight λ to construct a cost matrix: X+λF), the matching accuracy can be further improved at this time, but the speed is limited by the features The calculation time of information F is difficult to real-time

But its disadvantage is that prioritizing the matching of position-related information X will lead to wrong matching of position-related information X and affect the matching set of feature information F

That is to say, the calculation of the feature information F cannot correct the error caused by the position-related information X in the first ReID matching.

In this case, in multi-target tracking, especially in the case of overlapping and interleaving multiple targets, using position-related information X to match may make mistakes, and feature information F cannot be repaired

Images