Granted Patent B2

US 12,633,122 · App. 17/763,480 · Granted May 19, 2026

Inventor: Dezheng Zeng (Beijing, CN)

Assignees: Beijing Wodong Tianjun Information Technology Co., Ltd.; Beijing Jingdong Century Trading Co., Ltd.

G06V20/49G06V10/7715G06V20/46G09B5/065

Abstract

Claims (57)

1 . A video segmentation method, comprising:

acquiring a to-be-segmented video, and determining a correspondence between knowledge point data in the to-be-segmented video and video frames in the to-be-segmented video; and

segmenting the to-be-segmented video according to the correspondence to obtain at least one video segment;

wherein determining the correspondence between the knowledge point data in the to-be-segmented video and the video frames in the to-be-segmented video comprises:

inputting the to-be-segmented video into a pre-trained video segmentation model, acquiring segmentation data outputted from the video segmentation model, and determining a video frame number interval corresponding to the knowledge point data according to the segmentation data;

wherein segmenting the to-be-segmented video according to the correspondence to obtain the at least one video segment comprises:

determining blurry demarcation points of the knowledge point data according to the video frame number interval corresponding to the knowledge point data;

acquiring, based on the blurry demarcation points, candidate video frames within a set range, performing boundary detection on the candidate video frames, and obtaining target demarcation points corresponding to the knowledge point data; and

determining a video segment corresponding to the knowledge point data according to the target demarcation points corresponding to the knowledge point data.

2 . The method of claim 1 , after obtaining the at least one video segment, further comprising:

in a case where the at least one video segment is a plurality of video segments, determining an association relationship between the plurality of video segments according to an association relationship between the knowledge point data, and determining at least one learning path according to the association relationship between the plurality of video segments, wherein the at least one learning path is used for characterizing a learning order between the plurality of video segments.

3 . The method of claim 2 , before determining the association relationship between the plurality of video segments according to the association relationship between the knowledge point data, further comprising:

acquiring a knowledge graph, and determining the association relationship between the knowledge point data according to the knowledge graph.

4 . The method of claim 3 , wherein acquiring the knowledge graph comprises:

extracting the knowledge point data comprised in the to-be-segmented video; and

performing relationship extraction on the knowledge point data, and constructing the knowledge graph comprising the association relationship between the knowledge point data.

5 . The method of claim 2 , further comprising:

determining, in response to a detected video viewing instruction, a learning path corresponding to the video viewing instruction; and

generating path recommendation information according to the learning path, and sending the path recommendation information to a client for display.

6 . The method of claim 1 , further comprising:

acquiring a to-be-segmented sample video and segmentation data corresponding to the to-be-segmented sample video; and

generating training sample pairs based on the to-be-segmented sample video and the segmentation data corresponding to the to-be-segmented sample video, and training a pre-constructed video segmentation model by using the training sample pairs to obtain a trained video segmentation model.

7 . A computer device, comprising:

at least one processor; and a storage apparatus storing at least one program;

wherein the at least one program, when executed by the at least one processor, causes the at least one processor to perform:

acquiring a to-be-segmented video, and determining a correspondence between knowledge point data in the to-be-segmented video and video frames in the to-be-segmented video; and

segmenting the to-be-segmented video according to the correspondence to obtain at least one video segment;

wherein determining the correspondence between the knowledge point data in the to-be-segmented video and the video frames in the to-be-segmented video comprises:

inputting the to-be-segmented video into a pre-trained video segmentation model, acquiring segmentation data outputted from the video segmentation model, and determining a video frame number interval corresponding to the knowledge point data according to the segmentation data;

wherein segmenting the to-be-segmented video according to the correspondence to obtain the at least one video segment comprises:

determining blurry demarcation points of the knowledge point data according to the video frame number interval corresponding to the knowledge point data;

acquiring, based on the blurry demarcation points, candidate video frames within a set range, performing boundary detection on the candidate video frames, and obtaining target demarcation points corresponding to the knowledge point data; and

determining a video segment corresponding to the knowledge point data according to the target demarcation points corresponding to the knowledge point data.

8 . The computer device of claim 7 , after obtaining the at least one video segment, further performing:

in a case where the at least one video segment is a plurality of video segments, determining an association relationship between the plurality of video segments according to an association relationship between the knowledge point data, and determining at least one learning path according to the association relationship between the plurality of video segments, wherein the at least one learning path is used for characterizing a learning order between the plurality of video segments.

9 . The computer device of claim 8 , before determining the association relationship between the plurality of video segments according to the association relationship between the knowledge point data, further comprising:

acquiring a knowledge graph, and determining the association relationship between the knowledge point data according to the knowledge graph.

10 . The computer device of claim 9 , wherein acquiring the knowledge graph comprises:

extracting the knowledge point data comprised in the to-be-segmented video; and

performing relationship extraction on the knowledge point data, and constructing the knowledge graph comprising the association relationship between the knowledge point data.

11 . The computer device of claim 8 , further performing:

determining, in response to a detected video viewing instruction, a learning path corresponding to the video viewing instruction; and

generating path recommendation information according to the learning path, and sending the path recommendation information to a client for display.

12 . The computer device of claim 7 , further comprising:

acquiring a to-be-segmented sample video and segmentation data corresponding to the to-be-segmented sample video; and

generating training sample pairs based on the to-be-segmented sample video and the segmentation data corresponding to the to-be-segmented sample video, and training a pre-constructed video segmentation model by using the training sample pairs to obtain a trained video segmentation model.

13 . A non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs:

acquiring a to-be-segmented video, and determining a correspondence between knowledge point data in the to-be-segmented video and video frames in the to-be-segmented video; and

segmenting the to-be-segmented video according to the correspondence to obtain at least one video segment;

wherein determining the correspondence between the knowledge point data in the to-be-segmented video and the video frames in the to-be-segmented video comprises:

inputting the to-be-segmented video into a pre-trained video segmentation model, acquiring segmentation data outputted from the video segmentation model, and determining a video frame number interval corresponding to the knowledge point data according to the segmentation data;

wherein segmenting the to-be-segmented video according to the correspondence to obtain the at least one video segment comprises:

determining blurry demarcation points of the knowledge point data according to the video frame number interval corresponding to the knowledge point data;

acquiring, based on the blurry demarcation points, candidate video frames within a set range, performing boundary detection on the candidate video frames, and obtaining target demarcation points corresponding to the knowledge point data; and

determining a video segment corresponding to the knowledge point data according to the target demarcation points corresponding to the knowledge point data.

14 . The non-transitory computer-readable storage medium of claim 13 , after obtaining the at least one video segment, further performing:

in a case where the at least one video segment is a plurality of video segments, determining an association relationship between the plurality of video segments according to an association relationship between the knowledge point data, and determining at least one learning path according to the association relationship between the plurality of video segments, wherein the at least one learning path is used for characterizing a learning order between the plurality of video segments.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (20)