Object tracking method, tracking processing method, corresponding apparatus, and electronic device

Patent metadata

Field	Value
Publication number	US-11967089-B2
Application number	US-202117335910-A
Country	US
Kind code	B2
Filing date	Jun 1, 2021
Priority date	Apr 18, 2019
Publication date	Apr 23, 2024
Grant date	Apr 23, 2024

Abstract

Embodiments of this application provide an object tracking method performed by a computer device. The method includes, when a target object is lost in a second image frame in a first subsequent image frames, determining, according to a first local feature and in second subsequent image frames starting with the second image frame, a third image frame in which the target object reappears after the target object is lost during the tracking; determining a location of a target object region in the third image frame including the target object; and continuing to track the target object in image frames according to the location of the target object region in the third image frame. Through the object tracking method, a lost object can be detected and repositioned by using an extracted first local feature of the target object, thereby effectively resolving the problem in the existing technical solution.

First claim

What is claimed is:

1. An object tracking method performed by a computer device, the method comprising:
determining a kernelized correlation filter (KCF) algorithm parameter according to device information of the computer device;
initializing the KCF algorithm according to the KCF parameter;
extracting a first local feature of a target object in an initial target object region of a first image frame in a video stream;
tracking the target object in first subsequent image frames after the first image frame in the video stream by using the KCF algorithm; and
performing the following operations when the target object is lost in a second image frame in the first subsequent image frames during the tracking:
determining, according to the first local feature and in second subsequent image frames starting with the second image frame, a third image frame in which the target object reappears for the first time after the target object is lost during the tracking;
determining a location of a target object region in the third image frame including the target object; and
continuing to track the target object in image frames after the third image frame according to the location of the target object region in the third image frame.

2. The object tracking method according to claim 1 , wherein the determining, according to the first local feature and in second subsequent image frames starting with the second image frame, a third image frame in which the target object reappears for the first time after the target object is lost during the tracking comprises:
repeating, starting from the second image frame in the second subsequent image frames, the following operations for the second subsequent image frames:
extracting a second local feature in an image frame and matching the second local feature with the first local feature to obtain a matching result, until the matching result indicates a matching success, an image frame corresponding to the successful matching being the third image frame.

3. The object tracking method according to claim 2 , wherein the matching the second local feature with the first local feature to obtain a matching result comprises:
determining distances between each feature point in the second local feature and feature points in the first local feature;
determining feature point matching pairs between the second local feature with the first local feature according to the distances between the feature point in the second local feature and the feature points in the first local feature;
obtaining, when a quantity of the feature point matching pairs is greater than a first threshold, a matching result indicating a matching success; and
obtaining, when the quantity of the feature point matching pairs is not greater than the first threshold, a matching result indicating a matching failure.

4. The object tracking method according to claim 3 , wherein the determining feature point matching pairs between the second local feature and the first local feature according to the distances between the feature point in the second local feature and the feature points in the first local feature comprises:
determining, for the feature point in the second local feature, a ratio of a second minimum distance to a minimum distance in the distances between the feature point and the feature points in the first local feature; and
determining, for the feature point when the ratio corresponding to the feature point is greater than a second threshold, that the feature point and a feature point in the first local feature with a minimum distance from the feature point form a feature point matching pair.

5. The object tracking method according to claim 3 , wherein the determining a location of a target object region in the third image frame including the target object comprises:
determining a homography matrix between the third image frame and the first image frame according to the feature point matching pairs; and
determining, based on a location of the initial target object region in the first image frame and the homography matrix, the location of the target object region in the third image frame including the target object.

6. The object tracking method according to claim 1 , further comprising determining a tracking state of the target object in the first subsequent image frames in the following manners, the tracking state being tracking successful or tracking lost:
determining, for a fourth image frame in the first subsequent image frames according to a location of the target object region in a previous image frame of the fourth image frame, a target response value in the fourth image frame for the target object region in the previous image frame;
determining that the target object is lost during the tracking in the fourth image frame when the target response value is not greater than a third threshold; and
determining that the target object is tracked successfully in the fourth image frame when the target response value is greater than the third threshold.

7. The object tracking method according to claim 1 , wherein the KCF parameter is selected from candidate KCF parameters according to test frame rates corresponding to the candidate KCF parameters.

8. The object tracking method according to claim 1 , wherein the determining a KCF parameter according to device information comprises:
determining a device level according to the device information; and
determining the KCF parameter according to the device level.

9. The object tracking method according to claim 8 , wherein the determining a device level according to the device information comprises:
determining the device level according to a preset mapping table and the device information,
the preset mapping table comprising correspondences between device information and device levels.

10. The object tracking method according to claim 9 , wherein the KCF parameter comprises a KCF feature and a filter kernel, and the determining the KCF parameter according to the device level comprises:
determining a histogram of oriented gradients (HOG) feature as the KCF feature and a Gaussian kernel as the filter kernel when the device level is a first level;
determining a grayscale pixel as the KCF feature and a Gaussian kernel as the filter kernel when the device level is a second level; or
determining a grayscale pixel as the KCF feature and a linear kernel as the filter kernel when the device level is a third level,
a device at the first level outperforming a device at the second level, a device at the second level outperforming a device at the third level.

11. A computer device, comprising:
a memory, configured to store a plurality of programs; and
a processor, configured to perform, when executing the plurality of programs stored in the memory, a plurality of operations including:
determining a kernelized correlation filter (KCF) algorithm parameter according to device information of the computer device;
initializing the KCF algorithm according to the KCF parameter:
extracting a first local feature of a target object in an initial target object region of a first image frame in a video stream;
tracking the target object in first subsequent image frames after the first image frame in the video stream by using the KCF algorithm; and
performing the following operations when the target object is lost in a second image frame in the first subsequent image frames during the tracking:
determining, according to the first local feature and in second subsequent image frames starting with the second image frame, a third image frame in which the target object reappears for the first time after the target object is lost during the tracking;
dete

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Wang Yitong
• Huang Jun
• Ji Xing

Classifications

• G06T7/248 Primary

  involving reference images or patches · CPC title

• G06T7/73

  using feature-based methods · CPC title

• G06T2207/10016

  Video; Image sequence · CPC title

• G06T7/246 Primary

  using feature-based methods, e.g. the tracking of corners or segments · CPC title

• G06T2207/20081

  Training; Learning · CPC title