Template tracking method, device, augmented reality system and storage medium

Abstract

The invention proposes a template tracking method, device, augmented reality system and computer-readable storage medium. The template tracking method includes: setting a template picture in advance; matching the template picture in the current image in the continuously collected images, where the continuously collected images are collected by a camera; and using a tracking algorithm to calculate the template in the current image in the continuously collected images. The estimated pose of the picture; using a matching algorithm to calculate the precise pose of the template picture in the current image in the continuously collected images based on the estimated pose; and returning the estimated pose if the precise pose of the template picture in the current image is not calculated; in Returns the precise pose after calculating the precise pose of the template image in the current image. The embodiments of the present invention improve the tracking accuracy and speed, and can be applied to AR template image tracking, so that the model can closely adhere to the template image, allowing users to have a better AR experience; they have low hardware requirements, lower the usage threshold, and expand the user group.

Description

technical field [0001] The present invention relates to the field of information technology, in particular to a template tracking method, device, augmented reality system and computer-readable storage medium. Background technique [0002] There are two types of 2D tracking algorithms in the prior art. One is only a matching algorithm, that is, template matching is performed on each frame captured by the camera. The other is the tracking and matching algorithm, which usually has two synchronous threads, one for the tracking algorithm and the other for the matching algorithm. Input a picture, two threads will start at the same time, and then end at the same time, and finally combine the results of the two threads to obtain the pose of the template image in the current picture. [0003] The above two algorithms have the following defects: for only matching algorithms, the calculation speed is often slow, and the calculated poses between the two frames before and after are not...

AI Technical Summary

Problems solved by technology

[0003] The above two algorithms have the following defects: for only matching algorithms, the calculation speed is often slow, and the calculated poses between the two frames before and after are not smooth, which is not suitable for AR (Augmented Reality, augmented reality) model display; for synchronous The matching and tracking algorithm is limited by the performance of the mobile processor. Either use a fast but not very effective matching algorithm, which makes the tracking speed fast but the attitude accuracy is low, or use a slow matching algorithm with good matching effect. This way the attitude accuracy is high, but the speed is slow

Images