An intelligent processing method and system for training videos based on end-cloud cooperation

Abstract

The application provides a training video intelligent processing method and system based on end-cloud cooperation, and belongs to the technical field of video processing. The application obtains a training video segment sequence of a terminal and segment-associated training content item identifiers, content version identifiers and content association types, compares training content version data before and after updating, determines original version explanation segments and original version reference segments, and forms a replacement interval. The application sends candidate video segments and the replacement interval to the terminal, receives inconsistent boundaries and target content item identifiers determined by the terminal according to a candidate playing order, and updates interval boundaries according to segment positions. The application repeatedly updates interval boundaries for newly-included video segments, obtains a target replacement interval, and determines target video segments from a content-updated video and sends the target video segments to the terminal. The application can accurately determine a complete updating range in a stock video updating scene after training content is changed, and maintains the coherence and version consistency of training content before and after updating.

Description

Technical Field

[0001] This application belongs to the field of video processing technology, and in particular relates to a method and system for intelligent processing of training videos based on edge-cloud collaboration. Background Technology

[0002] With the development of digital learning platforms, government and enterprise training activities have generated a large number of training videos that require long-term storage, repeated use, and continuous maintenance. A cloud-edge collaborative intelligent processing method for training videos combines cloud processing capabilities with the video storage and playback capabilities of end devices, making it suitable for applications such as updating training content and maintaining existing videos.

[0003] Existing technologies typically convert instructional text into recognizable text data, identify the audio content in instructional videos, and determine video segments that do not meet the requirements by comparing the audio content with the instructional text. These non-compliant video segments are then deleted, and the remaining video segments are rearranged and synthesized.

[0004] In large-scale digital training scenarios for government and enterprises, changes to regulations, business processes, or course content often necessitate partial updates to training videos already published or stored on the terminal. The same training content may be scattered across multiple video segments and interconnected with preceding and following content. Existing methods of partial identification and replacement easily overlook content requiring simultaneous updates, resulting in coexistence of old and new content, incomplete contextual connections, or repeated modifications. Therefore, existing technologies present a technical challenge in accurately determining the complete scope of updates and maintaining consistency between the updated and pre-updated content after training content changes. Summary of the Invention

[0005] The purpose of this application is to provide a method, system, and electronic device for intelligent processing of training videos based on edge-cloud collaboration, so as to solve the problem in the prior art that it is difficult to accurately determine the complete scope of the update and maintain the consistency of the content before and after the update after the training content is changed.

[0006] To address the aforementioned technical problems, firstly, this application provides a method for intelligent processing of training videos based on edge-cloud collaboration, applied in the cloud, including: Obtain the training video segment sequence of the terminal. Each training video segment is associated with at least one set of content association records. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. The content association type is either a description type or a reference type. Based on the changed content item identifier, original content version identifier, and target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and reference type, the original version description fragment and the original version reference fragment are determined. The replacement interval is determined based on the original version description fragments arranged in chronological order. Candidate video fragments selected from the content update video and the replacement interval are sent to the terminal so that the terminal can generate inconsistency boundaries and target content item identifiers based on the candidate video fragments and the replacement interval. Update the start or end point of the replacement interval based on the original version description fragment or original version reference fragment corresponding to the inconsistency boundary and target content item identifier sent by the terminal. The target replacement interval is obtained by repeatedly updating the start or end point of the replacement interval, with the termination condition being that the original version description fragments or original version reference fragments corresponding to each content association type associated with the newly included training video fragments in the replacement interval after each update are all located within the updated replacement interval. The target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval are sent to the terminal.

[0007] In one feasible implementation, based on the changed content item identifier, the original content version identifier, and the target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and reference type, the original version description fragment and the original version reference fragment are determined, including: Compare the training content version data before and after the update according to the training content item identifier; When two content version identifiers for the same training content item are different, the training content item identifier is determined to be the changed content item identifier, and the content version identifiers before and after the update are determined to be the original content version identifier and the target content version identifier, respectively. Filter the original version content association records from the content association records to associate the changed content item identifier and the original content version identifier; The training video clips associated with the original version content association records, which are of the description type and the reference type respectively, are designated as the original version description clips and the original version reference clips.

[0008] In one feasible implementation, the replacement interval is determined based on the original version description fragments arranged in chronological order, including: Determine the start and end positions of each original version instruction segment within the training video segment sequence; In the original version description fragments arranged in chronological order of their starting positions, the starting position of the first original version description fragment is determined as the starting point of the interval, and the ending position of the last original version description fragment is determined as the ending point of the interval, thus obtaining the replacement interval.

[0009] In one feasible implementation, each candidate video segment is associated with at least one set of content association records; To enable the terminal to generate inconsistency boundaries and target content item identifiers based on candidate video segments and replacement intervals, including: Based on the replacement interval, the terminal replaces the training video segments within the replacement interval with candidate video segments in the training video segment sequence to generate a candidate playback order; The terminal determines the adjacent video segments located before and after the candidate video segments according to the candidate playback order, and obtains the boundary adjacent segment information including the candidate video segment and the adjacent video segments; Based on the information of adjacent segments at the boundary, the terminal determines the boundary between the candidate video segment that is associated with the same training content item identifier but has a different content version identifier and any adjacent video segment as an inconsistency boundary, and determines the training content item identifier that is associated with the candidate video segment and the adjacent video segment on both sides of each inconsistency boundary as the target content item identifier corresponding to the inconsistency boundary.

[0010] In one feasible implementation, the inconsistency boundary includes at least one of a start boundary and an end boundary; The starting boundary is the playback switching position between the adjacent video segment preceding the candidate video segment in the candidate playback order and the candidate video segment that is ranked first in the candidate video segment. The ending boundary is the playback switching position between the candidate video segment that is ranked last in the candidate video segment and the adjacent video segment that is ranked after the candidate video segment.

[0011] In one feasible implementation, the start or end point of the replacement interval is updated based on the original version description fragment or original version reference fragment corresponding to the inconsistency boundary and target content item identifier sent by the terminal, including: When the inconsistency boundary includes the starting boundary, the starting point of the replacement interval is updated to the target content item identifier corresponding to the associated starting boundary in the original version description fragment, and the starting position of the original version description fragment that is located before the replacement interval and last in the time sequence in the training video fragment sequence; When the inconsistency boundary includes an end boundary, the endpoint of the replacement interval is updated to the target content item identifier corresponding to the associated end boundary in the original version reference fragment, and the end position of the original version reference fragment located after the replacement interval and last in the time sequence in the training video fragment sequence.

[0012] In one feasible implementation, the termination condition is that the original version description fragments or original version reference fragments corresponding to each content association type associated with the newly included training video fragments in the replacement interval after each update are all located within the updated replacement interval. The target replacement interval is obtained by repeatedly updating the start or end point of the replacement interval, including: The newly added training video segments that are covered by the replacement range after each update are identified as newly added training video segments. When a newly added training video segment does not meet the termination condition, identify the video segments to be included that are outside the replacement interval and have the same training content item identifier associated with the newly added training video segment in the original version description segment or original version reference segment corresponding to each content association type associated with the newly added training video segment. Update the start and / or end points of the replacement interval according to the position of the video segments to be included in the training video segment sequence to obtain the updated replacement interval. Continue until the newly added training video segments in the updated replacement interval meet the termination condition to obtain the target replacement interval.

[0013] In one feasible implementation, the start and / or end points of the replacement interval are updated based on the position of the video segment to be included in the training video segment sequence to obtain the updated replacement interval, including: Determine the content association type for each group of content association records associated with the newly added training video clips; When the content association type is a reference type, the starting position of the original version description segment that is the last in the time sequence of the video segment to be included and is associated with the same training content item identifier before the replacement interval is determined as the starting point after the update. When the content association type is description type, the end position of the original version reference segment that is located after the replacement interval and is associated with the same training content item identifier in the video segment to be included is determined as the end position after the update. The updated replacement interval is determined based on the updated start and end points.

[0014] In one feasible implementation, sending the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal includes: Based on the time sequence of training video clips, determine the content association records and arrangement order of each training video clip within the target replacement interval, and replace the content version identifier in the content association record of the associated changed content item identifier and the corresponding original content version identifier with the corresponding target content version identifier to obtain the target content association record for each training video clip. Select video segments related to the target content from the updated video and arrange them in the order of arrangement to obtain the target video segments; Send the start and end points of the target replacement interval, along with the target video segment, to the terminal.

[0015] Secondly, this application provides a training video intelligent processing system based on edge-cloud collaboration, applied in the cloud, including: The acquisition module is used to acquire the training video segment sequence of the terminal. Each training video segment is associated with at least one set of content association records. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. The content association type is either a description type or a reference type. The determination module is used to determine the original version description fragment and the original version reference fragment based on the changed content item identifier, the original content version identifier, and the target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and reference type. The sending module is used to determine the replacement interval based on the original version description fragments arranged in chronological order, and to send the candidate video fragments and replacement intervals selected from the content update video to the terminal so that the terminal can generate inconsistency boundaries and target content item identifiers based on the candidate video fragments and replacement intervals. The update module is used to update the start or end point of the replacement interval based on the original version description fragment or original version reference fragment corresponding to the inconsistency boundary and target content item identifier sent by the terminal. The update module is also used to repeatedly update the start or end point of the replacement interval to obtain the target replacement interval, with the termination condition being that the original version description fragments or original version reference fragments corresponding to each content association type associated with the training video fragments newly included in the replacement interval after each update are all located within the updated replacement interval. The sending module is also used to send the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal.

[0016] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute computer programs to implement the steps of the edge-cloud collaborative training video intelligent processing method as described in the first aspect above.

[0017] The intelligent training video processing method based on edge-cloud collaboration provided in this application establishes associations between training video segments and training content items, content versions, and content association types. When training content changes, it identifies explanatory and referenced segments related to the original content version and forms replacement intervals accordingly. The terminal identifies inconsistencies between candidate and retained video segments according to the actual playback order, allowing the cloud to adjust the replacement intervals based on the location of the inconsistent content. By further identifying associated content and expanding the intervals for newly included video segments, relevant video segments requiring synchronous updates are included in the complete update scope, thereby determining the video segments corresponding to the target content version from the updated content video. This reduces the coexistence of old and new training content, incomplete contextual connections, and repeated modifications. Therefore, this application can accurately determine the complete update scope in scenarios involving partial updates of existing videos after training content changes, while maintaining the continuity and version consistency of the training content before and after the update. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating an intelligent training video processing method based on edge-cloud collaboration, provided for an embodiment of this application; Figure 2 A schematic diagram of edge-cloud interaction for an intelligent training video processing method based on edge-cloud collaboration provided in an embodiment of this application; Figure 3 A flowchart illustrating a cyclic update and replacement interval method provided in an embodiment of this application; Figure 4 A schematic diagram of the structure of a training video intelligent processing system based on edge-cloud collaboration provided in an embodiment of this application; Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] To address the problems of existing technologies, embodiments of this application provide a method, apparatus, device, computer storage medium, and computer program product for intelligent processing of training videos based on edge-cloud collaboration. The method for intelligent processing of training videos based on edge-cloud collaboration provided in this application embodiment will be described first below.

[0022] This application applies to scenarios involving the continuous maintenance of existing training videos within large-scale government and enterprise digital learning platforms. In application environments where training content is uniformly managed in the cloud and training videos are stored or cached on terminals, changes to regulations, business processes, product knowledge, or course content can lead to the same training content being scattered across different video segments and referenced multiple times in subsequent segments. When making partial updates to related training videos, issues such as incomplete update scope, the coexistence of old and new content, and inconsistencies in the connection between different training sessions can easily arise.

[0023] This application includes a cloud server and at least one terminal device. The terminal device can be a desktop computer, laptop computer, tablet computer, smartphone, conference display terminal, or training all-in-one machine, used for storing, caching, and playing training videos; the cloud server is used to manage training content version data and content update videos. The terminal device communicates with the cloud server via a wired or wireless network.

[0024] Figure 1 This illustration shows a flowchart of an embodiment of the intelligent processing method for training videos based on edge-cloud collaboration provided in this application. Figure 1 As shown, the method includes steps S110 to S160.

[0025] Figure 2 This illustration shows a schematic diagram of cloud-terminal data interaction provided in one embodiment of this application. For example... Figure 2 As shown, during the processing steps S110 to S160, the terminal sends the training video segment sequence and content association record to the cloud, the cloud sends candidate video segments and replacement intervals to the terminal, the terminal sends the inconsistency boundary and corresponding target content item identifier to the cloud, and the cloud sends the target replacement interval and target video segment to the terminal.

[0026] S110. Obtain the training video segment sequence of the terminal. Each training video segment is associated with at least one set of content association records. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. The content association type is either a description type or a reference type.

[0027] In large-scale government and enterprise digital training scenarios, after receiving a data acquisition request from the cloud, the terminal reads the locally saved or cached training video playback list and forms a training video segment sequence S=[P1, P2, ..., Pn] based on the sequence number or segment start time. Here, Pi represents the training video segment at position i in the sequence, where i is any integer from 1 to n, and n represents the number of training video segments. Each training video segment Pi is associated with a segment identifier, sequence number, start time, end time, and a content association record set Ri. This content association record set can be represented as Ri={Ri1, Ri2, ..., Rim}, where m represents the number of content association records associated with training video segment Pi. Any content association record Rij represents the j-th group of content association records associated with the i-th training video segment Pi, where j is any integer from 1 to m. The content association record Rij can be represented as Rij=[Cij, Vij, Tij], where Cij is the training content item identifier, Vij is the content version identifier, and Tij is the content association type.

[0028] Content association records can be generated during the production or release of training videos based on the training content items in the course catalog, the content version recorded in the training script, and the scope of explanation corresponding to the video segment. These records are saved along with the playback information of the corresponding training video segment. Training content item identifiers are used to distinguish between policy clauses, business processes, product knowledge points, or course conclusions, and remain unchanged when the same training content item undergoes version changes. Content version identifiers can be generated based on the release batch or update order of the corresponding training content item. The content association type Tij can be either a descriptive type or a reference type. A descriptive type is used when the training video segment directly explains the corresponding training content item, while a reference type is used when explaining other content and using the corresponding training content item as a premise or background.

[0029] For example, the training video segment sequence is S=[P1, P2, P3, P4]. P1 directly explains version V1 of training content item Ca, corresponding to R1={[Ca, V1, description type]}; P2 directly explains version V1 of training content item Cb and references version V1 of training content item Ca, corresponding to R2={[Cb, V1, description type], [Ca, V1, reference type]}; P3 directly explains version V1 of training content item Cc and references version V1 of training content item Cb, corresponding to R3={[Cc, V1, description type], [Cb, V1, reference type]}; P4 references version V1 of training content item Cc, corresponding to R4={[Cc, V1, reference type]}. Ca, Cb, and Cc are the specific values ​​of the training content item identifier Cij, respectively.

[0030] like Figure 2 As shown, the terminal sends a data group D composed of the training video segment sequence S, the segment identifier, sequence number, start time, end time of each training video segment, and content association record sets R1 to Rn, to the cloud. The training video segment sequence S and the content association record sets R1 to Rn in data group D correspond to... Figure 2 The training video clip sequence and content association record.

[0031] S120. Based on the changed content item identifier, original content version identifier, and target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and reference type, determine the original version description fragment and the original version reference fragment.

[0032] The previous training content version data record shows the training content items and versions used in the current training video, while the updated training content version data record the training content items and versions after the training content was adjusted. By analyzing the content version changes of the same training content item in the two datasets, the changed content item and its corresponding version are determined. Then, by combining this with the content association type in the content association record, the corresponding training video segment is identified.

[0033] In one feasible implementation, step S120 determines the original version description fragment and the original version reference fragment based on the changed content item identifier, the original content version identifier, and the target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and reference type, including: Compare the training content version data before and after the update based on the training content item identifier.

[0034] The training content version data before and after the update uses the same data structure. Each data entry includes a training content item identifier and a content version identifier, and uses the same identifier encoding method as the content-associated record. The training content version data before the update can be represented as B0, and the training content version data after the update can be represented as B1. Each data entry in B0 and B1 can be represented as [C, V], where C represents the training content item identifier and V represents the content version identifier.

[0035] Specifically, B0 is read from the pre-update course release data stored in the cloud, and B1 is read from the post-update course release data. Data indexes are then created for each index, using the training content item identifier as the search key. Based on the training content item identifier in B0, data with the same training content item identifier is searched in B1. The training content item identifier, the pre-update content version identifier, and the post-update content version identifier from the two found data entries are sequentially combined to form a version comparison record. For example, B0 = {[Ca, V1], [Cb, V1], [Cc, V1]}, B1 = {[Ca, V2], [Cb, V2], [Cc, V2]}. After searching for Ca, Cb, and Cc, the version comparison record set J = {[Ca, V1, V2], [Cb, V1, V2], [Cc, V1, V2]} is obtained.

[0036] When two versions of the same training content item have different identifiers, the training content item identifier is determined to be the changed content item identifier, and the content version identifiers before and after the update are determined to be the original content version identifier and the target content version identifier, respectively.

[0037] The changed content item identifier is determined by comparing two content version identifiers in the same version comparison record. The first content version identifier in each version comparison record comes from the training content version data before the update, and the second content version identifier comes from the training content version data after the update, so they can be used as the original content version identifier and the target content version identifier, respectively.

[0038] Specifically, the version comparison records in the version comparison record set J are read sequentially, and the consistency of two content version identifiers in the same version comparison record is compared. When the two content version identifiers are different, the corresponding training content item identifier and the two content version identifiers are retained to form a changed version record. Based on the version comparison record set J, the changed version record set H = {[Ca, V1, V2], [Cb, V1, V2], [Cc, V1, V2]} can be obtained, where Ca, Cb, and Cc are the changed content item identifiers, V1 in each changed version record is the original content version identifier, and V2 is the target content version identifier.

[0039] Filter the original version content association records from the content association records to find the associated changed content item identifier and the original content version identifier.

[0040] Original version content association records refer to content association records that simultaneously associate the changed content item identifier and the corresponding original content version identifier. Original version content association records retain the content association type and the relationship between the record and the training video clip, so as to determine the corresponding training video clip based on the description type and reference type.

[0041] Specifically, the content association records in data group D, from set R1 to Rn, are read sequentially. The training content item identifier and content version identifier in each set of content association records are then compared with the changed content item identifier and original content version identifier of each changed version record in set H. When both identifiers are identical, the corresponding content association record is identified as the original version content association record. Based on the example in S110, original version content association records R11, R21, R22, R31, R32, and R41 can be obtained. Each original version content association record retains its original content association type and associated training video clip.

[0042] The training video clips associated with the original version content association records, which are of the description type and the reference type respectively, are designated as the original version description clips and the original version reference clips.

[0043] An original version description fragment refers to a training video fragment associated with an original version content association record when the content association type is description. An original version reference fragment refers to a training video fragment associated with an original version content association record when the content association type is reference. When a training video fragment is associated with multiple sets of original version content association records, and the content association types of the different original version content association records are different, the same training video fragment can simultaneously serve as both an original version description fragment and an original version reference fragment for different training content items.

[0044] Specifically, the content association types in each original version content association record are read sequentially. When the content association type is a description type, the corresponding training video segment is identified as the original version description segment corresponding to the corresponding changed content item identifier; when the content association type is a reference type, the corresponding training video segment is identified as the original version reference segment corresponding to the corresponding changed content item identifier. Thus, the original version description segment corresponding to training content item Ca is determined to be P1, and the original version reference segment is determined to be P2; the original version description segment corresponding to training content item Cb is determined to be P2, and the original version reference segment is determined to be P3; the original version description segment corresponding to training content item Cc is determined to be P3, and the original version reference segment is determined to be P4, resulting in the original version description segment set {P1, P2, P3} and the original version reference segment set {P2, P3, P4}.

[0045] S130. Determine the replacement interval based on the original version description fragments arranged in chronological order, and send the candidate video fragments selected from the content update video and the replacement interval to the terminal so that the terminal can generate inconsistency boundaries and target content item identifiers based on the candidate video fragments and the replacement interval.

[0046] The updated content video is a video created based on the updated training content and stored in the cloud. The candidate video clip is a video clip selected from the updated content video and used to temporarily replace training video clips within the specified range.

[0047] In one feasible implementation, step S130, which determines the replacement interval based on the original version description fragments arranged in chronological order, includes: Determine the start and end positions of each original version instruction segment within the training video segment sequence.

[0048] The start and end positions of the original version's instructional segment represent the start and end times of that segment on the training video playback timeline, respectively. All training video segments use the same time base and time unit for their start and end positions, which can be obtained from the start and end times of the corresponding training video segments in data group D.

[0049] Specifically, based on the segment identifier of the original version instruction segment, the start and end times of the corresponding training video segment are retrieved from data group D, forming a location record set L. Each location record in the location record set L can be represented as [Pi, Ai, Ei], where Pi represents the original version instruction segment, Ai represents the start position of the original version instruction segment Pi, and Ei represents the end position of the original version instruction segment Pi. Based on the original version instruction segment set {P1, P2, P3} determined in S120, the location record set L = {[P1, A1, E1], [P2, A2, E2], [P3, A3, E3]} can be obtained.

[0050] In the original version description fragments arranged in chronological order of their starting positions, the starting position of the first original version description fragment is determined as the starting point of the interval, and the ending position of the last original version description fragment is determined as the ending point of the interval, thus obtaining the replacement interval.

[0051] Arranging the original version description clips in chronological order of their starting positions means using the starting position in each record as the basis for arrangement, placing the original version description clip with the earliest starting position first. The original version description clips at the beginning and end of the arrangement represent the earliest and latest starting positions, respectively. The replacement interval represents the continuous playback range defined by the start and end points of the interval.

[0052] Specifically, the position records in the position record set L are arranged from earliest to latest according to their starting positions. Based on the playback order in the training video segment sequence S, A1 precedes A2, and A2 precedes A3, resulting in the arrangement [P1, P2, P3]. The starting position A1 of P1 is determined as the interval start point, and the ending position E3 of P3 is determined as the interval end point, resulting in the replacement interval K0 = [A1, E3]. The replacement interval K0 corresponds to training video segments P1 to P3.

[0053] In one feasible implementation, each candidate video segment is associated with at least one set of content association records.

[0054] The content association records associated with candidate video segments use the same data structure as the content association records in S110. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. Updated videos can be divided into video segment sequences W=[Q1, Q2, Q3, Q4] according to the scope of the training content, where Q1 to Q4 represent video segments created according to the updated training content. Q1 is associated with {[Ca, V2, description type]}, Q2 with {[Cb, V2, description type], [Ca, V2, reference type]}, Q3 with {[Cc, V2, description type], [Cb, V2, reference type]}, and Q4 with {[Cc, V2, reference type]}.

[0055] Specifically, the content association records associated with training video segments P1 to P3 within the replacement interval K0 are read, and each content association record is mapped to the set of changed version records H. For content association records that associate changed content item identifiers with original content version identifiers, the content version identifiers are adjusted to the corresponding target content version identifiers, while keeping the training content item identifier and content association type unchanged, thus obtaining the adjusted content association record sets corresponding to P1 to P3 respectively. Using each adjusted content association record set as a matching condition, video segments with the same training content item identifier, content version identifier, and content association type are searched from the video segment sequence W, determining Q1, Q2, and Q3 as candidate video segments corresponding to the replacement interval K0, forming the candidate video segment sequence Q=[Q1, Q2, Q3]. Figure 2 As shown, the cloud sends the candidate video segment sequence Q and the replacement interval K0 to the terminal. The candidate video segment sequence Q and the replacement interval K0 correspond to... Figure 2 Candidate video segments and replacement intervals.

[0056] Step S130, used by the terminal to generate inconsistency boundaries and target content item identifiers based on candidate video segments and replacement intervals, includes: Based on the replacement interval, the terminal replaces the training video segments within the replacement interval with candidate video segments in the training video segment sequence, generating a candidate playback order.

[0057] The candidate playback order refers to the playback arrangement formed after the terminal temporarily places candidate video segments into the replacement interval without changing the actual stored content of the local training video segments. The candidate playback order retains the original relative playback order of each training video segment outside the replacement interval and places the candidate video segments according to the order of arrangement in the candidate video segment sequence.

[0058] Specifically, based on the start point A1 and end point E3 of the replacement interval K0, the terminal locates P1 to P3 in the training video segment sequence S, removes the playback records corresponding to P1 to P3 from the temporary playlist, and then sequentially inserts Q1, Q2, and Q3 from the position corresponding to the start point of the interval. P4 is located after the replacement interval K0 and maintains its original playback position, thereby generating the candidate playback order Sc=[Q1, Q2, Q3, P4].

[0059] The terminal determines the adjacent video segments located before and after the candidate video segments according to the candidate playback order, and obtains the boundary adjacent segment information including the candidate video segment and the adjacent video segments.

[0060] A retained video segment refers to a video segment in the candidate playback order that is outside the replacement interval and continues to use the original video content from the training video segment sequence S. An adjacent video segment refers to a retained video segment that is immediately adjacent to the beginning or end of the candidate video segment sequence in the candidate playback order. Boundary adjacent segment information includes the segment identifier, arrangement order, and content association record of the candidate video segment, as well as the segment identifier, arrangement order, and content association record of the adjacent video segments located before and after the candidate video segment.

[0061] Specifically, the terminal locates the first candidate video segment Q1 and the last candidate video segment Q3 in the candidate playback order Sc, and reads the adjacent playback records before Q1 and after Q3. In the current example, there are no reserved video segments before Q1, and the adjacent video segment after Q3 is P4. The terminal reads the candidate video segment sequence Q and the content association record associated with P4 to form the boundary adjacent segment information N. The boundary adjacent segment information N includes the content association records from Q1 to Q3 and the content association record associated with P4 {[Cc, V1, reference type]}.

[0062] Based on the information of adjacent segments at the boundary, the terminal determines the boundary between the candidate video segment that is associated with the same training content item identifier but has a different content version identifier and any adjacent video segment as an inconsistency boundary, and determines the training content item identifier that is associated with the candidate video segment and the adjacent video segment on both sides of each inconsistency boundary as the target content item identifier corresponding to the inconsistency boundary.

[0063] Inconsistency boundaries are determined by comparing the content association records of candidate video segments with their corresponding adjacent video segments. The comparison first performs equivalence matching based on the training content item identifier, then compares the consistency of the content version identifiers in the two sets of content association records with the same training content item identifier. The target content item identifier is the training content item identifier that is commonly associated with both sets of content association records but has different content version identifiers.

[0064] Specifically, at the end of the candidate video segment sequence, the terminal compares the content association records associated with Q3 and P4 group by group. The content association record [Cc, V2, description type] associated with Q3 and the content association record [Cc, V1, reference type] associated with P4 have the same training content item identifier Cc and different content version identifiers V2 and V1. Therefore, the playback switch position between Q3 and P4 is determined as the inconsistency boundary, and Cc is determined as the target content item identifier corresponding to the inconsistency boundary. The content association record [Cb, V2, reference type] associated with Q3 and the content association record [Cc, V1, reference type] associated with P4 have different training content item identifiers and do not form an inconsistency boundary.

[0065] In one feasible implementation, the inconsistency boundary includes at least one of a start boundary and an end boundary.

[0066] The start boundary and end boundary correspond to the beginning and end of the candidate video segment sequence, respectively. Inconsistency boundaries may include only the start boundary, only the end boundary, or both the start and end boundaries. Each inconsistency boundary retains the corresponding target content item identifier.

[0067] Specifically, the terminal checks the beginning and end of the candidate video segment sequence. For the beginning, the playback transition position between the first candidate video segment and its preceding adjacent video segment is used as the starting boundary to be checked; for the end, the playback transition position between the last candidate video segment and its following adjacent video segment is used as the ending boundary to be checked. In the current example, the candidate video segment sequence starts from the beginning of the training video segment sequence, without forming adjacent segment pairs with starting boundaries, and an ending boundary is formed between Q3 and P4.

[0068] The starting boundary is the playback switching position between the adjacent video segment preceding the candidate video segment in the candidate playback order and the candidate video segment that is ranked first in the candidate video segment. The ending boundary is the playback switching position between the candidate video segment that is ranked last in the candidate video segment and the adjacent video segment that is ranked after the candidate video segment.

[0069] The playback transition position indicates the sequential position in the candidate playback order where the previous video segment finishes playing and the playback transitions to the next video segment. The playback transition position can be represented by the boundary type, the video segment identifiers on both sides of the boundary, and the target content item identifier, to distinguish the content version differences that exist at the two ends of the candidate video segment sequence.

[0070] Specifically, based on the adjacent arrangement of Q3 and P4 in the candidate playback order Sc, the terminal represents the detection result as [end boundary, segment identifier of Q3, segment identifier of P4, Cc]. Here, the end boundary indicates the type of inconsistency boundary, the segment identifiers of Q3 and P4 represent the video segments on either side of the playback switch position, and Cc represents the corresponding target content item identifier. For example... Figure 2 As shown, the terminal sends the detection results to the cloud. The end boundary and Cc in the detection results correspond to... Figure 2 Inconsistency boundaries and corresponding target content item identifiers.

[0071] S140. Update the start or end point of the replacement interval according to the original version description fragment or original version reference fragment corresponding to the inconsistency boundary and target content item identifier sent by the terminal.

[0072] In one feasible implementation, step S140 updates the start or end point of the replacement interval based on the original version description fragment or original version reference fragment corresponding to the inconsistency boundary and target content item identifier sent by the terminal, including: When the inconsistency boundary includes the starting boundary, the starting point of the replacement interval is updated to the target content item identifier corresponding to the associated starting boundary in the original version description fragment, and the starting position of the original version description fragment located before the replacement interval and last in the time sequence in the training video fragment sequence.

[0073] The original version description fragment preceding the replacement interval refers to the original version description fragment arranged in the training video fragment sequence before the corresponding training video fragment at the start of the interval. The original version description fragment at the end of the time sequence refers to the training video fragment with the latest starting position among the original version description fragments that meet the aforementioned positional conditions. The training content item identifier associated with the original version description fragment is the same as the target content item identifier corresponding to the starting boundary.

[0074] Specifically, the boundary detection results sent by the terminal are read. When the boundary detection results include the starting boundary, the target content item identifier corresponding to the starting boundary is extracted from the boundary detection results, and training video segments with the same training content item identifier are searched from the original version description segments determined in S120. According to the arrangement number in data group D, the original version description segments located before the current replacement interval are filtered, and then arranged from latest to earliest according to their starting positions. The starting position of the first original version description segment is read, and the read starting position replaces the starting point of the current interval, while the ending point of the current interval remains unchanged. In the current example, the boundary detection results do not include the starting boundary, so the starting point of the replacement interval K0 remains A1.

[0075] When the inconsistency boundary includes an end boundary, the endpoint of the replacement interval is updated to the target content item identifier corresponding to the associated end boundary in the original version reference fragment, and the end position of the original version reference fragment located after the replacement interval and last in the time sequence in the training video fragment sequence.

[0076] The original version reference fragment following the replacement interval refers to the original version reference fragment arranged in the training video fragment sequence after the training video fragment corresponding to the end of the interval. The original version reference fragment at the end of the time sequence refers to the training video fragment with the latest ending position among the original version reference fragments that meet the aforementioned position conditions. The training content item identifier associated with the original version reference fragment is the same as the target content item identifier corresponding to the end boundary.

[0077] Specifically, when the boundary detection result includes an end boundary, the target content item identifier corresponding to the end boundary is extracted from the boundary detection result, and training video clips with the same training content item identifier are searched from the original version reference clips determined in S120. Based on the permutation number in data group D, the original version reference clips located after the current replacement interval are filtered, and then arranged from latest to earliest according to their end positions. The end position of the first original version reference clip is read, and the read end position replaces the end point of the current interval, while the start point of the current interval remains unchanged.

[0078] In the current example, the boundary detection result is [end boundary, fragment identifier of Q3, fragment identifier of P4, Cc]. The target content item identifier Cc is extracted from the boundary detection result, and the original version reference fragment P4 associated with the training content item identifier Cc is found. Based on the permutation number in data group D, P4 is determined to be located after the replacement interval K0, and the end position E4 of P4 is read. Since P4 is the original version reference fragment located after the replacement interval K0 and last in time sequence, the interval endpoint is updated from E3 to E4, while keeping the interval start point A1 unchanged, resulting in the updated replacement interval K1 = [A1, E4]. When the boundary detection result includes both a start boundary and an end boundary, new interval start and end points are determined according to the target content item identifiers corresponding to the two boundaries, and the updated replacement interval is determined by the new interval start and end points together.

[0079] S150. The target replacement interval is obtained by repeatedly updating the start or end point of the replacement interval, with the termination condition being that the original version description fragments or original version reference fragments corresponding to each content association type associated with the training video fragments newly included in the replacement interval after each update are all located within the updated replacement interval.

[0080] In step S150, the content association types associated with the newly added training video clips refer to the content association types that the content association records associated with the changed content item identifier and the corresponding original content version identifier in the newly added training video clips have. Content association records that are not associated with both the changed content item identifier and the corresponding original content version identifier are not included in the termination condition judgment.

[0081] In one feasible implementation, such as Figure 3 As shown, step S150 specifically includes steps S151 to S155.

[0082] S151. The newly added training video segments covered by the replacement interval after each update are identified as newly added training video segments.

[0083] Newly added training video segments refer to the training video segments that are covered by the replacement range after this update compared to the replacement range before the update. Newly added training video segments do not include those already within the replacement range before the update. This can be determined by comparing the start and end points of the range before and after the update, as well as the sequence number of the training video segments.

[0084] Specifically, the start and end points of the interval before and after the update are saved respectively. When the interval start point moves forward, the training video segments to be added and covered between the new interval start point and the original interval start point are read from data group D; when the interval end point moves backward, the training video segments to be added and covered between the original interval end point and the new interval end point are read, and the read training video segments are identified as newly added training video segments. In the current example, the interval is updated from K0=[A1, E3] to K1=[A1, E4], and the newly covered training video segment is P4. Therefore, the set of newly added training video segments F1={P4} is obtained, and the set of content associated records R4={[Cc, V1, reference type]} associated with P4 is read.

[0085] When a newly added training video segment does not meet the termination condition, identify the video segments to be included that are outside the replacement interval and have the same training content item identifier associated with the newly added training video segment in the original version description segment or original version reference segment corresponding to each content association type associated with the newly added training video segment. Update the start and / or end points of the replacement interval according to the position of the video segments to be included in the training video segment sequence to obtain the updated replacement interval. Continue until the newly added training video segments in the updated replacement interval meet the termination condition to obtain the target replacement interval.

[0086] S152. For each content association record in the newly added training video clip that is associated with the changed content item identifier and the corresponding original content version identifier, whether the original version description fragment or the original version reference fragment corresponding to the content association type is located within the updated replacement range.

[0087] Termination conditions are determined based on the content association type of each group of content association records in the newly added training video clips. For content association records of the reference type, the original version description fragment that is associated with the same training content item identifier, is located before the corresponding reference fragment, and is the last in the time sequence is taken as the corresponding original version description fragment; for content association records of the description type, the original version reference fragment that is associated with the same training content item identifier, is located after the corresponding description fragment, and is the last in the time sequence is taken as the corresponding original version reference fragment.

[0088] If the original version description fragment or original version reference fragment corresponding to each content association type is located within the updated replacement range, the judgment result of step S152 is yes, and step S155 is executed; if there is a corresponding fragment located outside the updated replacement range, the judgment result of step S152 is no, and step S153 is executed.

[0089] Specifically, the content association records associated with newly added training video clips are read sequentially. For reference-type content association records, the earliest original version description clip is searched according to the training content item identifier in chronological order; for description-type content association records, the latest original version reference clip is searched according to the training content item identifier in chronological order. Based on the permutation number, start position, and end position in data group D, it is determined whether the found training video clip is located within the current replacement interval.

[0090] S153. Identify the original version description fragments or original version reference fragments that are outside the replacement interval and have the same training content item identifier associated with the newly added training video fragments as video fragments to be included.

[0091] Specifically, the original version description fragments or original version reference fragments that are determined in step S152 to be located outside the updated replacement interval and have the same training content item identifier associated with the newly added training video fragments are added to the video fragment set M to be included. If the video fragment set M to be included is not empty, step S154 is executed.

[0092] S154. Update the start and / or end points of the replacement interval based on the location of the video segment to be included.

[0093] In one feasible implementation, step S154 updates the start and / or end points of the replacement interval based on the position of the video segment to be included in the training video segment sequence, including: Determine the content association type for each set of content association records associated with the newly added training video clips.

[0094] Determining the content association type involves reading the content association type field from each set of content association records associated with the newly added training video clip, and performing an equivalence match between the read field values ​​and the description type and reference type. To maintain the correspondence between the newly added training video clip, the training content item identifier, and the content association type, the read results can be compiled into association check records. Each association check record includes the clip identifier of the newly added training video clip, the training content item identifier, and the content association type.

[0095] Specifically, the process iterates through the newly added training video clips in the set and the content association records associated with each new training video clip, reading the training content item identifier and content association type to form an association check record set. In the current example, the newly added training video clip P4 is associated with the content association record [Cc, V1, reference type], thus forming an association check record set Z1 = {[P4, Cc, reference type]}. The association check record set Z1 is used to find the original version description clip corresponding to the training content item identifier Cc.

[0096] When the content association type is a reference type, the starting position of the original version description segment that is the last in the time sequence of the video segment to be included and is associated with the same training content item identifier before the replacement interval is determined as the starting point of the update.

[0097] The video clips to be included for the reference type refer to the original version description clips that have the same training content item identifier as the corresponding associated inspection record, are located before the corresponding newly added training video clips, and have not yet been included in the current replacement interval. The original version description clip at the end of the time sequence represents the training video clip with the latest starting position among the original version description clips that meet the aforementioned conditions.

[0098] Specifically, for association check records with a reference type of content association, the corresponding training video segment is searched from the original version description segments according to the training content item identifier, and the original version description segments that precede the corresponding newly added training video segment are filtered according to the sorting number in data group D. The filtered results are sorted from latest to earliest according to the starting position, and the original version description segment that is ranked first and located before the current replacement interval is added to the set of video segments to be included M, and the corresponding starting position is determined as the starting point of the updated interval.

[0099] When the content association type is description type, the end position of the original version reference segment that is located after the replacement interval and is associated with the same training content item identifier in the video segment to be included is determined as the end point after the update.

[0100] The video clips to be included corresponding to the description type refer to the original version reference clips that have the same training content item identifier as the corresponding associated inspection record, are located after the corresponding newly added training video clips, and have not yet been included in the current replacement interval. The original version reference clip at the end of the time sequence refers to the training video clip with the latest ending position among the original version reference clips that meet the aforementioned conditions.

[0101] Specifically, for association check records with the content association type of "description", the corresponding training video segment is searched from the original version reference segments according to the training content item identifier, and the original version reference segments located after the corresponding newly added training video segment are filtered according to the sorting number in data group D. The filtered results are sorted from latest to earliest according to the end position, and the original version reference segment ranked first and located after the current replacement interval is added to the set of video segments to be included M, and the corresponding end position is determined as the end point of the updated interval.

[0102] The updated replacement interval is determined based on the updated start and end points.

[0103] The updated replacement range is defined by both the updated start and end points. When only the start point is updated, the end point before the update is retained; when only the end point is updated, the start point before the update is retained; when both the start and end points are updated, the two updated positions form a new continuous playback range.

[0104] Specifically, the replacement interval after this update is determined based on the updated start and end points of the intervals obtained from each associated check record, and the interval range before and after this update is compared to redetermine the new training video segment. After this update is completed, the updated replacement interval is used as the current replacement interval and the process returns to step S151.

[0105] S155. The replacement interval that meets the termination condition is determined as the target replacement interval.

[0106] When the judgment result of step S152 is yes, the replacement interval is no longer updated, and the current replacement interval that meets the termination condition is determined as the target replacement interval. In the current example, a new training video clip P4 is associated with a content association record of the reference type. The original version description clip P3 corresponding to P4 is already located in the replacement interval K1, and P4 is not associated with a content association record of the description type. Therefore, there is no video clip to be included. The judgment result of step S152 is yes, steps S153 and S154 are not executed, and K1=[A1, E4] is determined as the target replacement interval.

[0107] S160. Send the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal.

[0108] The target video segment is the video segment that matches the content association record with the corresponding target content association record within the target replacement interval.

[0109] In one feasible implementation, step S160, which sends the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal, includes: Based on the chronological order of the training video clip sequence, determine the content association records and arrangement order of each training video clip within the target replacement interval. Replace the content version identifier in the content association records of the associated changed content item identifier and the corresponding original content version identifier with the corresponding target content version identifier to obtain the target content association records corresponding to each training video clip.

[0110] The target content association record represents the training content item identifier, content version identifier, and content association type corresponding to the training video segment within the target replacement interval when the target content version is used. For content association records that associate both the changed content item identifier and the original content version identifier, the original content version identifier is replaced with the target content version identifier corresponding to the same changed content item identifier in the changed version record set H; for content association records that do not associate both the changed content item identifier and the original content version identifier, the original record content remains unchanged.

[0111] Specifically, based on the target replacement interval K1=[A1, E4], the training video segment P1 corresponding to the interval start point A1 and the training video segment P4 corresponding to the interval end point E4 are located from the training video segment sequence S. P1 to P4 and their corresponding content association record sets R1 to R4 are read in chronological order. Using the training content item identifier and content version identifier in each set of content association records as search conditions, the changed version record set H is searched for records where the changed content item identifier and the original content version identifier are the same. If the search is successful, V1 in the content association record is replaced with the corresponding V2, while keeping the training content item identifier and content association type unchanged; if no corresponding changed version record is found, the corresponding content association record remains unchanged. This yields the target content associated record sets Y1 to Y4, where Y1 = {[Ca, V2, description type]}, Y2 = {[Cb, V2, description type], [Ca, V2, reference type]}, Y3 = {[Cc, V2, description type], [Cb, V2, reference type]}, and Y4 = {[Cc, V2, reference type]}.

[0112] Select video segments related to the target content from the updated video and arrange them in the order of arrangement to obtain the target video segments.

[0113] Video clips associated with target content records refer to video clips whose training content item identifier, content version identifier, and content association type are identical to the corresponding target content association record. Target video clips are arranged in chronological order of the corresponding training content within the target replacement interval.

[0114] Specifically, using the target content association record sets Y1 to Y4 as matching conditions, the content association records associated with each video segment in the video segment sequence W of the content update video are read sequentially. Equivalence matching is performed on the training content item identifier, content version identifier, and content association type to determine that Y1 corresponds to Q1, Y2 to Q2, Y3 to Q3, and Y4 to Q4. Following the order of P1, P2, P3, and P4 within the target replacement interval K1, Q1, Q2, Q3, and Q4 are arranged to obtain the target video segment sequence Qt=[Q1, Q2, Q3, Q4]. Q1 to Q3 are the candidate video segments determined in step S130, and Q4 is the video segment added after the target replacement interval is expanded to P4.

[0115] Send the start and end points of the target replacement interval, along with the target video segment, to the terminal.

[0116] The data sent to the terminal includes the start and end points of the target replacement interval, as well as the target video segments arranged in the order of the training content within the target replacement interval. The start and end points of the interval use the same time reference as the start and end positions of the training video segments in data group D, which is used to locate the continuous playback range that needs to be replaced in the current training video segment sequence on the terminal.

[0117] Specifically, the video update data U=[A1, E4, Qt] is composed of the starting point A1, the ending point E4 of the target replacement interval K1, and the target video segment sequence Qt. Each target video segment in the target video segment sequence Qt carries a segment identifier, video data, and a corresponding target content association record. For example... Figure 2 As shown, the cloud sends the video update data U to the terminal, with A1 and E4 corresponding to it. Figure 2 The target replacement interval in Qt corresponds to... Figure 2 The target video segment is located in A1 and E4. The terminal locates the target and replaces the corresponding training video segment in the interval, and processes Q1 to Q4 according to the order in Qt.

[0118] After receiving the video update data U, the terminal locates the training video segments P1 to P4 to be replaced in the current training video segment sequence based on the interval start point A1 and interval end point E4. It retains the training video segments before and after the target replacement interval and their corresponding playback order, and places the target video segment sequence Qt into the corresponding position of the target replacement interval. The terminal associates the target content association records carried by Q1 to Q4 with the corresponding target video segments, re-determines the sequence number, start time, and end time based on the playback duration of each target video segment, forming an updated training video segment sequence. The updated training video segment sequence and the target video segments are then saved to local storage or cache.

[0119] The intelligent training video processing method based on edge-cloud collaboration provided in this application determines the initial replacement range by comparing training content versions in the cloud and determining the initial replacement range based on explanatory and referential relationships. The terminal checks the version consistency between candidate video segments and retained video segments in conjunction with the local playback order. Then, the range boundary is repeatedly expanded according to the content association relationship of the newly added training video segments, which can completely include explanatory and referential segments related to the changed content. On this basis, the target video segments are determined and distributed according to the target replacement range, which can reduce content omissions, coexistence of old and new versions, and inconsistencies in the connection between the preceding and following content when partially updating existing training videos, thereby improving the accuracy of the training video update scope and the coherence of the updated training content.

[0120] Figure 4 This is a schematic diagram illustrating a specific implementation of an edge-cloud collaborative training video intelligent processing system provided in this application. (Refer to...) Figure 4 Applied to the cloud, the system may include: The acquisition module 410 is used to acquire a sequence of training video segments from the terminal. Each training video segment is associated with at least one set of content association records. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. The content association type is either a description type or a reference type. The determination module 420 is used to determine the original version description fragment and the original version reference fragment based on the changed content item identifier, the original content version identifier and the target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and the reference type. The sending module 430 is used to determine the replacement interval based on the original version description fragments arranged in chronological order, and to send the candidate video fragments selected from the content update video and the replacement interval to the terminal so that the terminal can generate inconsistency boundaries and target content item identifiers based on the candidate video fragments and the replacement interval. The update module 440 is used to update the start or end point of the replacement interval based on the original version description fragment or original version reference fragment corresponding to the inconsistency boundary and target content item identifier sent by the terminal. The update module 440 is also used to repeatedly update the start or end point of the replacement interval to obtain the target replacement interval, with the termination condition being that the original version description fragments or original version reference fragments corresponding to each content association type associated with the training video fragments newly included in the replacement interval after each update are all located within the updated replacement interval. The sending module 430 is also used to send the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal.

[0121] The edge-cloud collaborative training video intelligent processing system of this application embodiment is used to implement the aforementioned edge-cloud collaborative training video intelligent processing method. Therefore, the specific implementation of the edge-cloud collaborative training video intelligent processing system can be found in the embodiment section of the edge-cloud collaborative training video intelligent processing method above. The specific implementation can be referred to the description of the corresponding embodiment, and will not be repeated here.

[0122] Figure 5 A schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application is shown.

[0123] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0124] Specifically, the processor 510 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0125] Memory 520 may include mass storage for data or instructions. For example, and not as a limitation, memory 520 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these.

[0126] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices.

[0127] The processor 510 reads and executes computer program instructions stored in the memory 520 to implement any of the edge-cloud collaborative training video intelligent processing methods in the above embodiments.

[0128] In one example, the electronic device may also include a communication interface 530 and a bus 540. Wherein, such as Figure 5 As shown, the processor 510, memory 520, and communication interface 530 are connected through bus 540 and complete communication with each other.

[0129] The communication interface 530 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0130] Bus 540 includes hardware, software, or both, which couples the components of an electronic device together.

[0131] The electronic device can execute the intelligent training video processing method based on edge-cloud collaboration in the embodiments of this application, thereby realizing the intelligent training video processing method based on edge-cloud collaboration described in conjunction with the accompanying drawings.

[0132] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0133] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave.

[0134] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0135] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0136] The above provides a detailed description of the intelligent processing method, system, and electronic device for training videos based on edge-cloud collaboration provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for intelligent processing of training videos based on edge-cloud collaboration, applied in the cloud, characterized in that, include: Obtain a sequence of training video clips from the terminal. Each training video clip is associated with at least one set of content association records. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. The content association type is either a description type or a reference type. Based on the changed content item identifier, original content version identifier, and target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and the reference type, the original version description fragment and the original version reference fragment are determined. The replacement interval is determined based on the original version description fragments arranged in chronological order. Candidate video fragments selected from the content update video and the replacement interval are sent to the terminal so that the terminal can generate inconsistency boundaries and target content item identifiers based on the candidate video fragments and the replacement interval. Update the start or end point of the replacement interval based on the inconsistency boundary and the original version description fragment or original version reference fragment corresponding to the target content item identifier sent by the terminal. The target replacement interval is obtained by repeatedly updating the start or end point of the replacement interval, with the termination condition being that the original version description fragments or original version reference fragments corresponding to each content association type associated with the training video fragments newly included in the replacement interval after each update are all located within the updated replacement interval. The target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval are sent to the terminal.

2. The method according to claim 1, characterized in that, The step of determining the original version description fragment and the original version reference fragment based on the changed content item identifier, the original content version identifier, and the target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and the reference type, includes: Compare the training content version data before the update with the training content version data after the update according to the training content item identifier; When two content version identifiers of the same training content item are different, the training content item identifier is determined to be a changed content item identifier, and the content version identifiers before and after the update are determined to be the original content version identifier and the target content version identifier, respectively. Filter the original version content association records that associate the changed content item identifier and the original content version identifier from the content association records; The training video clips associated with the original version content association records, whose content association types are description type and reference type respectively, are designated as original version description clips and original version reference clips.

3. The method according to claim 1, characterized in that, The process of determining the replacement interval based on the original version description fragments arranged in chronological order includes: Determine the start and end positions of each original version instruction segment within the sequence of training video segments; In the original version description fragments arranged in the order of their starting positions, the starting position of the first original version description fragment is determined as the interval start point, and the ending position of the last original version description fragment is determined as the interval end point, thus obtaining the replacement interval.

4. The method according to claim 1, characterized in that, Each candidate video segment is associated with at least one set of content association records; The method for generating inconsistency boundaries and target content item identifiers for the terminal based on candidate video segments and replacement intervals includes: Based on the replacement interval, the terminal replaces the training video segments within the replacement interval with the candidate video segments in the training video segment sequence to generate a candidate playback order; The terminal determines the adjacent video segments located before and after the candidate video segments according to the candidate playback order, and obtains boundary adjacent segment information including the candidate video segment and the adjacent video segments; Based on the boundary adjacent segment information, the terminal determines the boundary between the candidate video segment that is associated with the same training content item identifier but has a different content version identifier and any adjacent video segment as the inconsistency boundary, and determines the training content item identifier that is commonly associated with the candidate video segment and the adjacent video segment located on both sides of each inconsistency boundary as the target content item identifier corresponding to the inconsistency boundary.

5. The method according to claim 4, characterized in that, The inconsistency boundary includes at least one of a start boundary and an end boundary; The starting boundary is the playback switching position between the adjacent video segment preceding the candidate video segment in the candidate playback order and the candidate video segment that is ranked first in the candidate video segment, and the ending boundary is the playback switching position between the candidate video segment that is ranked last in the candidate video segment and the adjacent video segment following the candidate video segment.

6. The method according to claim 5, characterized in that, The step of updating the start or end point of the replacement interval based on the inconsistency boundary and the original version description fragment or original version reference fragment corresponding to the target content item identifier sent by the terminal includes: When the inconsistency boundary includes the starting boundary, the starting point of the replacement interval is updated to the target content item identifier associated with the starting boundary in the original version description fragment, and the starting position of the original version description fragment located before the replacement interval and last in the time sequence in the training video fragment sequence; When the inconsistency boundary includes the end boundary, the endpoint of the replacement interval is updated to the target content item identifier associated with the end boundary in the original version reference fragment, and the end position of the original version reference fragment located after the replacement interval and last in time sequence in the training video fragment sequence.

7. The method according to claim 1, characterized in that, The termination condition is that the original version description fragments or original version reference fragments corresponding to each content association type associated with the training video fragments newly included in the replacement interval after each update are all located within the updated replacement interval. The process of repeatedly updating the start or end point of the replacement interval to obtain the target replacement interval includes: The newly added training video segments covered by the replacement interval after each update are identified as newly added training video segments. When the newly added training video segment does not meet the termination condition, determine the video segments to be included that are outside the replacement interval and have the same training content item identifier associated with the newly added training video segment in the original version description segment or the original version reference segment corresponding to each content association type associated with the newly added training video segment. Update the start and / or end points of the replacement interval according to the position of the video segments to be included in the training video segment sequence to obtain the updated replacement interval. Continue until the newly added training video segments in the updated replacement interval meet the termination condition to obtain the target replacement interval.

8. The method according to claim 7, characterized in that, The step of updating the start and / or end points of the replacement interval based on the position of the video segment to be included in the training video segment sequence to obtain the updated replacement interval includes: Determine the content association type of each group of content association records associated with the newly added training video clip; When the content association type is the reference type, the starting position of the original version description segment that is the last in the time sequence of the video segment to be included, which is located before the replacement interval and is associated with the same training content item identifier, is determined as the updated starting point; When the content association type is the description type, the end position of the original version reference segment that is located after the replacement interval and associated with the same training content item identifier in the video segment to be included is determined as the end point after the update. The updated replacement interval is determined based on the updated start point and the updated end point.

9. The method according to claim 1, characterized in that, Sending the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal includes: According to the time sequence of the training video segments, determine the content association records and arrangement order of each training video segment within the target replacement interval, and replace the content version identifier in the content association record that associates the changed content item identifier and the corresponding original content version identifier with the corresponding target content version identifier to obtain the target content association record corresponding to each training video segment; The target video segment is obtained by selecting video segments associated with the target content association record from the content update video and arranging them in the specified order. The start and end points of the target replacement interval, as well as the target video segment, are sent to the terminal.

10. A training video intelligent processing system based on edge-cloud collaboration, applied in the cloud, characterized in that, include: The acquisition module is used to acquire a sequence of training video clips from the terminal. Each training video clip is associated with at least one set of content association records. Each set of content association records includes a training content item identifier, a content version identifier, and a content association type. The content association type is either a description type or a reference type. The determination module is used to determine the original version description fragment and the original version reference fragment based on the changed content item identifier, the original content version identifier, and the target content version identifier determined by comparing the training content version data before and after the update, as well as the description type and the reference type. The sending module is used to determine the replacement interval based on the original version description fragments arranged in chronological order, and to send the candidate video fragments selected from the content update video and the replacement interval to the terminal so that the terminal can generate an inconsistency boundary and a target content item identifier based on the candidate video fragments and the replacement interval. The update module is used to update the start or end point of the replacement interval based on the inconsistency boundary and the original version description fragment or original version reference fragment corresponding to the target content item identifier sent by the terminal. The update module is also used to repeatedly update the start or end point of the replacement interval to obtain the target replacement interval, with the termination condition being that the original version description fragments or original version reference fragments corresponding to each content association type associated with the training video fragments newly included in the replacement interval after each update are all located within the updated replacement interval. The sending module is also used to send the target replacement interval and the target video segment determined in the content-updated video based on the target replacement interval to the terminal.

Images