Target tracking method, device and storage medium based on human-computer interaction

Abstract

The present disclosure provides a target tracking method, device, and storage medium based on human-computer interaction. The target to be tracked can be determined through the trigger command generated based on the user's target selection operation, without the need for continuous multi-frame images. Operational learning can determine the target to be tracked, reduce target acquisition operations, improve target acquisition efficiency and tracking efficiency; and detect the position and scale of the target image in the next frame based on the extracted image features, to a certain extent Optimizing the tracking effects of rigid targets, scale-changing targets, high-speed moving targets, and long-term occluded targets, and improving the accuracy of target tracking; thus, it can also ensure computing efficiency and target tracking on general portable devices. Effect.

Description

technical field [0001] The present disclosure relates to the technical field of tracking, and in particular to a target tracking method, device and storage medium that can be applied to portable devices based on human-computer interaction. Background technique [0002] Tracking technology has always been one of the hot spots in the field of computer vision, and it has broad application prospects in many aspects such as military frame inspection, precision guidance, fire strikes, battlefield assessment, and security monitoring. [0003] At this stage, there are two methods for tracking moving target objects. The first one is to use KCF (Kernal Correlation Filter, kernel correlation filter algorithm) to superimpose HOG (Histogram of Oriented Gradient, histogram of oriented gradient) features to achieve tracking. Tracking of moving target objects; the second is to use KCF to superimpose depth features to realize the detection of moving target objects. When the first solution i...

AI Technical Summary

Problems solved by technology

When the first solution is adopted, when the target object is in a relatively complex scene, it is impossible to accurately locate and detect the target object, especially the ability to distinguish the foreground and background of the scene is not strong

Although the second solution can solve the problem of the first solution, its computational complexity is high, and it cannot run faster on portable devices with average computing power.

Images