Method and apparatus with object tracking using dynamic field of view

Patent metadata

Field	Value
Publication number	US-12456205-B2
Application number	US-202117541753-A
Country	US
Kind code	B2
Filing date	Dec 3, 2021
Priority date	Dec 4, 2020
Publication date	Oct 28, 2025
Grant date	Oct 28, 2025

Abstract

A method with object tracking includes: determining a first target tracking state by tracking a target from a first image frame with a first field of view (FoV); determining a second FoV based on the first FoV and the first target tracking state; and generating a second target tracking result by tracking a target from a second image frame with the second FoV.

First claim

What is claimed is:

1 . A method with object tracking, the method comprising:
determining a first target tracking state indicating whether a tracking state of a target that has been detected is stable by tracking the target from a first image frame using a first search region of the first image frame, the first search region having a first region size; determining, based on a result of the determining of the first target tracking state and the first region size, a second region size for a second search region of a second image frame from among a plurality of predetermined reference region sizes; and
generating a second target tracking result by tracking a target from the second image frame by using the second search region of the second image frame, the second search region having the second region size,
wherein the first region size and the second region size are selected from a first reference region size, a second reference region size, and a third reference region size, among the plurality of predetermined reference region sizes,
wherein the second reference region size is greater than the first reference region size,
wherein the third reference region size is less than the first reference region size, and
wherein the determining of the second region size comprises:
when the first region size is the first reference region size, determining the second region size as the second reference region size in response to the first target tracking state being a second state, determining the second region size as the third reference region size in response to the first target tracking state being a first state, and determining the second region size as the first reference region size in response to the first target tracking state being a third state;
when the first region size is the second reference region size, determining the second region size as the second reference region size in response to the first target tracking state being the second state, determining the second region size as the first reference region size in response to the first target tracking state being the first state, and determining the second region size as the second reference region size in response to the first target tracking state being the third state; and
when the first region size is the third reference region size, determining the second region size as the first reference region size in response to the first target tracking state being the second state, determining the second region size as the third reference region size in response to the first target tracking state being the first state, and determining the second region size as the third reference region size in response to the first target tracking state being the third state.

2 . The method of claim 1 , wherein the first image frame and the second image frame are collected by different image collectors of a same electronic device.

3 . The method of claim 1 , wherein the determining of the first target tracking state comprises:
generating a first target tracking result by tracking the target from the first image frame by using the first search region of the first image frame; and determining the first target tracking state based on the first target tracking result.

4 . The method of claim 3 , wherein the first target tracking result comprises a prediction confidence, and
wherein the determining of the first target tracking state based on the first target tracking result comprises determining the first target tracking state according to a result of a comparison of the prediction confidence and a preset threshold.

5 . The method of claim 4 , wherein the preset threshold comprises a first threshold and a second threshold, and
wherein the determining of the first target tracking state according to the result of the comparison comprises:
in response to the prediction confidence being greater than the first threshold, determining the first target tracking state as the first state;
in response to the prediction confidence being less than the second threshold, determining the first target tracking state as the second state; and
in response to the prediction confidence being greater than the second threshold and less than the first threshold, determining the first target tracking state as the third state.

6 . The method of claim 1 , wherein the generating of the second target tracking result comprises:
setting a reference template feature based on an initial image frame of an image sequence to which the second image frame belongs; determining the second search region of the second image frame based on the second region size and a position of the target of the first image frame, and obtaining a search feature from the second search region of the second image frame; and generating the second target tracking result based on the reference template feature and the search feature.

7 . The method of claim 6 , wherein the generating of the second target tracking result based on the reference template feature and the search feature comprises:
in response to the second region size being the second reference region size or the third reference region size, generating a scaled template feature by scaling the reference template feature; and generating the second target tracking result based on the scaled template feature and the search feature.

8 . The method of claim 7 , wherein the generating of the scaled template feature comprises:
in response to the second region size being the second reference region size, generating the scaled template feature by scaling down the reference template feature; and in response to the second region size being the third reference region size, generating the scaled template feature by scaling up the reference template feature.

9 . The method of claim 8 , wherein the generating of the second target tracking result based on the scaled template feature and the search feature comprises:
generating a feature map by performing a convolution operation on the scaled template feature and the search feature using the scaled template feature as a convolution kernel; and obtaining the second target tracking result based on the feature map.

10 . The method of claim 3 , wherein the first target tracking result comprises a prediction confidence, and either one or both of a target position and a target size, and
wherein the determining of the first target tracking state based on the first target tracking result comprises:
obtaining at least one of a target relative displacement corresponding to the first image frame and a ratio between a size of the target of the first image frame and a size of the first image frame, based on either one or both of the target position and the target size; and
determining the first target tracking state based on a result of a comparison of the prediction confidence and a preset threshold, and either one or both of a result of a comparison of the target relative displacement and a reference displacement and a result of a comparison of the ratio and a reference ratio.

11 . The method of claim 10 , wherein the determining of the first target tracking state based on the result of the comparison of the prediction confidence and the preset threshold, and either one or both of the result of the comparison of the target relative displacement and the reference displacement and the result of the comparison of the ratio and the reference ratio comprise:
in response to the prediction confidence being greater than a first threshold and the target relative displacement being less than the reference displacement, determining the first target tracking state as the first state; in respons

Assignees

• Samsung Electronics Co Ltd

Inventors

• Xu Jingtao
• Chen Yiwei
• Qiu Feng
• Yoo Byung In
• Park Changbeom
• Lee Hyunjeong
• Han Jaejoon
• Wang Qiang
• Yu Jiaqian

Classifications

• G06V10/40

  Extraction of image or video features · CPC title

• G06V10/22

  by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition · CPC title

• G06T7/292

  Multi-camera tracking · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2207/20081

  Training; Learning · CPC title