Granted Patent B2

US 12,462,521 · App. 18/195,914 · Granted Nov 4, 2025

Inventor: Kang-Hsi Liu (New Taipei, TW)

Assignee: Nanning FuLian FuGui Precision Industrial Co., Ltd.

G06V10/28G06F3/013G06V10/22

Abstract

Claims (46)

1 . A target image extraction method, comprising following steps:

obtaining operation behavior data of a user;

obtaining a specified coordinate based on the operation behavior data and obtaining a background screenshot based on the specified coordinate;

extracting a target image from the background screenshot, wherein the target image is provided with a target object and wherein extracting the target image from the background screenshot further comprises:

obtaining multiple candidate images based on the background screenshot;

sorting the candidate images based on image information and sequentially selecting one candidate image as a predicted target, wherein the image information includes an image entropy, and sorting the candidate images based on image information and sequentially selecting one candidate image as a predicted targets further comprises:

selecting the candidate image with highest image entropy as the predicted target;

performing image recognition in the background screenshot based on each predicted target;

determining whether the selected predicted target is recognized in the background screenshot and recognized only once;

using the predicted target with the highest image entropy and recognized only once in the background screenshot as the target image;

setting a target coordinate based on the target image; and

returning the target image and the target coordinate.

2 . The method of claim 1 , wherein the operation behavior data is provided with eye-tracking operation data.

3 . The method of claim 1 , wherein the operation behavior data is provided with mouse operation data.

4 . The method of claim 1 , wherein obtaining the specified coordinate based on the operation behavior data and obtaining a background screenshot based on the specified coordinate further comprises:

obtaining an image within a default range from the specified coordinate as the background screenshot.

5 . The method of claim 1 , wherein setting the target coordinate based on the target image further comprises:

setting a centroid of the target image as the target coordinate.

6 . The method of claim 1 , wherein extracting the target image from the background screenshot further comprises:

performing preprocessing on the candidate images, wherein the preprocessing is provided with edge extraction.

7 . A device for target image extraction, comprising:

a processor; and

a storage unit for storing a computer program, wherein when the computer program is executed by the processor, the processor is configured for:

obtaining operation behavior data from a user;

obtaining a specified coordinate based on the operation behavior data and obtaining a background screenshot based on the specified coordinate;

extracting a target image from the background screenshot, wherein the target image is provided with a target object and wherein extracting the target image from the background screenshot further comprises:

obtaining multiple candidate images based on the background screenshot;

sorting the candidate images based on image information and sequentially selecting one candidate image as a predicted target, wherein the image information includes an image entropy, and sorting the candidate images based on image information and sequentially selecting one candidate image as a predicted targets further comprises:

selecting the candidate image with highest image entropy as the predicted target;

performing image recognition in the background screenshot based on each predicted target;

determining whether the selected predicted target is recognized in the background screenshot and recognized only once;

using the predicted target with the highest image entropy and recognized only once in the background screenshot as the target image;

setting a target coordinate based on the target image; and

returning the target image and the target coordinate.

8 . A non-transitory computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is configured for:

obtaining operation behavior data from a user;

obtaining a specified coordinate based on the operation behavior data and obtaining a background screenshot based on the specified coordinate;

extracting a target image from the background screenshot, wherein the target image is provided with a target object and wherein extracting the target image from the background screenshot further comprises:

obtaining multiple candidate images based on the background screenshot;

sorting the candidate images based on image information and sequentially selecting one candidate image as a predicted target, wherein the image information includes an image entropy, and sorting the candidate images based on image information and sequentially selecting one candidate image as a predicted targets further comprises:

selecting the candidate image with highest image entropy as the predicted target;

performing image recognition in the background screenshot based on each predicted target;

determining whether the selected predicted target is recognized in the background screenshot and recognized only once;

using the predicted target with the highest image entropy and recognized only once in the background screenshot as the target image;

setting a target coordinate based on the target image; and

returning the target image and the target coordinate.

Assignments (1)

Continuity (1)

References Cited (11)