Short video stitching and distributing method, program and storage medium

Abstract

The invention relates to a short video stitching and distributing method, a program and a storage medium, the method comprises the following steps: constructing an intelligent layered material library, the intelligent layered material library comprises a six-dimensional material management system, which is respectively a frame picture material, a background video material, a dynamic patch material, a background music material, a transition special effect material and a dynamic text material; the dynamic combination duplicate removal engine realizes intelligent combination and duplicate removal of each material in the intelligent layered material library to generate a short video, and the dynamic combination duplicate removal engine adopts a double-factor duplicate removal mechanism of a frequency control module and a content detection module to ensure that the generated short video is unique in intelligent combination and content presentation of each material; and the intelligent distribution control system completes high-efficiency compliance distribution of short video contents through a multi-stage state circulation and real-time decision-making mechanism. Through large-scale commercial verification, the generation amount of a single template per hour is increased from 80 to 380, and the material reuse rate is reduced from 47% to 8.5%; the image content similarity is reduced from 61.2% to 4.3%, and the audio repetition rate is reduced from 58.7% to 2.1%.

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet short video production, and particularly relates to a method, program and storage medium for short video splicing and distribution based on dynamic material combination and intelligent distribution mechanism, and is particularly applicable to the rapid generation and copyright compliance management of promotional short videos in high-frequency concurrent scenarios. Background Art

[0002] The current short video production system has the following technical defects:

[0003] 1. Contradiction between generation efficiency and repetition rate: The repetition rate of video content generated by traditional template systems exceeds the detection threshold of mainstream platform algorithms, generally > 60%, resulting in 35% of the daily video content triggering the platform's duplicate content detection mechanism and being throttled;

[0004] 2. Copyright conflicts of multiple materials: When commercial materials are reused among multiple users, there is a lack of effective control mechanisms. Industry investigation reports show that 78% of short video copyright disputes stem from repeated calls of materials;

[0005] 3. Limitations of existing technical solutions: (1) Dynamic splicing solutions do not solve the problem of duplicate material combination, resulting in a high risk of high similarity in the generated content; (2) The intelligent distribution system lacks a real-time copyright verification module and cannot prevent the illegal dissemination of unauthorized materials; (3) Traditional rendering engines have insufficient support for multi-layer material synthesis, and the generation of complex templates takes a long time (> 15 seconds), exceeding the acceptable range of commercial scenarios. Summary of the Invention

[0006] The purpose of the present invention is to provide a method, program and storage medium for short video splicing and distribution in order to solve the contradictions between generation efficiency and repetition rate, multiple material copyright conflicts and limitations of existing technical solutions in the prior art; the present invention effectively solves the balance problem between video generation efficiency and content repetition rate under high-concurrency requirements, real-time copyright tracking and usage frequency control of multi-source materials, and rapid rendering and cross-platform compatibility guarantee of complex special effect videos.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A method for short video splicing and distribution, characterized by comprising the following steps:

[0009] Construct an intelligent hierarchical material library, and the intelligent hierarchical material library includes a material management system in six dimensions, namely frame picture materials, background video materials, dynamic patch materials, background music materials, transition effect materials and dynamic text materials;

[0010] The intelligent combination and duplicate elimination of each material in the intelligent hierarchical material library are realized through a dynamic combination duplicate elimination engine. The dynamic combination duplicate elimination engine adopts a dual-factor duplicate elimination mechanism of a frequency control module and a content detection module to ensure the uniqueness of the intelligent combination of materials and content presentation in the generated short videos;

[0011] The intelligent distribution control system completes the efficient and compliant distribution of short video content through a multi-level state transition and real-time decision-making mechanism.

[0012] The six dimensions of the intelligent hierarchical material library include:

[0013] The first dimension is frame picture materials, which are processed by RGB color gamut standardization with a resolution not lower than 1080P, and invisible digital watermarks are embedded combined with SHA-256 hash value verification;

[0014] The second dimension is background video materials, which adopt the H.265 / HEVC encoding format, the key frame interval is controlled within 0.5 seconds, and the copyright management is carried out based on the blockchain deposit and proof technology of Hyperledger Fabric, and each video is bound with an independent usage counter;

[0015] The third dimension is dynamic patch materials, the dynamic adjustment range of transparency is 20%-80%, the size ratio does not exceed 30% of the main picture area, and UUID unique identification binding and device fingerprint double verification are adopted;

[0016] The fourth dimension is background music materials, the sampling rate is fixed at 44.1kHz, the bit rate is not lower than 128kbps, and cross-platform authorization tracking is realized through audio fingerprint recognition technology;

[0017] The fifth dimension is transition effect materials, the effect duration range is 0.2-1.5 seconds, multi-layer Alpha channel superposition is supported, and discrete recombination technology is adopted to prevent the reuse of effect templates;

[0018] The sixth dimension is dynamic text materials, the font library is limited to the commercial authorization white list, the text displacement path is parametrically controlled, and the text content hash value is matched with the sensitive word blacklist database in real time.

[0019] The dynamic combination duplicate elimination engine adopts a dual-factor duplicate elimination mechanism of a frequency control module and a content detection module to ensure the uniqueness of material combination and content presentation in the generated short videos. Specifically:

[0020] The frequency control module dynamically tracks the call times and time intervals of each material to ensure that the reuse interval of elements within the same material group ≥ the preset threshold; the sliding window counting algorithm is used to count the usage of each material in the recent N generations;

[0021] The content detection module, based on multi-modal feature extraction technology, performs hash encoding and similarity comparison on each material combination, including the extraction and comparison of visual features, audio features, and text features. The visual features specifically refer to the extraction of PHash and color histograms of frame pictures or videos. The audio features specifically refer to the generation of Mel spectrogram fingerprints of background music. The text features specifically refer to the TF-IDF vectorization after word segmentation of dynamic text content;

[0022] The deduplication process logic includes steps of material pre-selection and frequency verification, combined hash generation, historical record comparison, content difference enhancement, and record update and locking;

[0023] The exception handling and degradation strategy include candidate exhaustion handling, real-time performance guarantee, and threshold dynamic adjustment.

[0024] The intelligent distribution control system includes four core processing stages and three auxiliary control modules:

[0025] The four core processing stages include to-be-generated queue management, real-time copyright verification, priority distribution execution, and post-distribution processing;

[0026] The three auxiliary control modules include an exception rejection processing module, a timeliness control module, and a data feedback module;

[0027] Efficient and compliant distribution of short video content is achieved through the four core processing stages and the three auxiliary control modules.

[0028] The to-be-generated queue management is as follows: The short videos generated by the dynamic deduplication engine enter the to-be-distributed queue according to the preset priority rules. The default policy of the priority rules is: timeliness > copyright level > generation timestamp. The short video formats in the to-be-distributed queue are standardized, metadata is injected, and the capacity of the distribution queue is controlled. The capacity control uses the token bucket algorithm to limit the length of the distribution queue. When the threshold is exceeded, an automatic expansion or degradation strategy is triggered.

[0029] The real-time copyright verification is as follows: Query the authorization status of each material in the short video through blockchain nodes for material-level verification; Compare the text content of the sensitive word library in real time and the audio fingerprints for similarity detection with the copyright music library. The materials in the short video that pass the real-time copyright verification are marked as the "distributable" state, and the abnormal materials are transferred to the manual review queue.

[0030] The priority distribution execution adopts a dynamic scheduling algorithm, specifically: calculate the basic weight by adding the weight coefficients of whether the user is a VIP, the linearly calculated value of the content validity period ratio, and the 24-hour distribution success rate. The weight coefficient of VIP is 0.4, and that of non-VIP is 0. The weight coefficient of the linearly calculated value of the content validity period ratio is 0 - 0.3. The weight coefficient of the 24-hour distribution success rate > 95% is 0.4, otherwise it is 0; perform priority distribution according to the magnitude of the basic weight value; dynamically update the weight coefficients every 5 minutes according to the network load and the platform interface status.

[0031] The post-distribution processing includes: immediately triggering the counter increment operation for the materials in the successfully distributed short videos to update the count. If the material is for commercial use, it will be automatically marked as "disabled" after the cumulative usage reaches the authorization limit; set the survival period for each material, and the materials that have not been distributed after the timeout will be automatically transferred to cold storage and archived to different storage levels according to the popularity grading.

[0032] The exception rejection processing module: for short videos with failed copyright verification, incompatible platform formats, or / and content compliance risks, automatically analyze the rejection reasons and generate repair suggestions. For items that can be automatically repaired, start the transcoding service, and after repair, add the short videos back to the head of the distribution queue; transfer the items that require manual intervention to the review work order system and notify the operation personnel.

[0033] The aging control module adopts a multi-level aging strategy, specifically: set a countdown for urgent content, and trigger a text message alarm if it is not distributed after the timeout; adopt a dynamic TTL mechanism for regular content and automatically adjust according to the queue pressure; enable the background silent distribution mode for long-tail content.

[0034] The data feedback module: real-time monitor the distribution success rate, average response time, and copyright dispute rate, adaptively optimize the LSTM prediction model trained based on historical data, and give early warnings of potential bottlenecks; generate a distribution quality report every hour and automatically optimize the parameters of the priority algorithm.

[0035] A program for short video splicing and distribution, characterized in that it includes computer-executable instructions, and when the instructions are executed by a processor, it realizes the method of short video splicing and distribution described above.

[0036] A storage medium, on which a computer program is stored, characterized in that when the computer program is executed by a processor, it realizes the method of short video splicing and distribution described above.

[0037] The beneficial effects of the present invention

[0038] Verified by large-scale commercial applications:

[0039] 1. Generation efficiency: The hourly generation volume of a single template has been increased from 80 in the traditional solution to 380.

[0040] 2. Duplication rate control: The similarity of the screen content is reduced from 61.2% in the traditional solution to 4.3%, and the audio duplication rate is reduced from 58.7% in the traditional solution to 2.1%.

[0041] 3. Copyright compliance: The reuse rate of materials is reduced from 47% to 8.5%;

[0042] 4. Duplication elimination efficiency: The average time-consuming for a single duplication elimination check is reduced from 45 ms in the traditional solution to 8 ms;

[0043] 5. Duplication rate control: Under the generation volume of tens of millions, the content duplication rate is stable below 3.2%, while it is 58 - 72% in the traditional solution.

[0044] 6. Resource consumption: The memory occupancy is reduced by 42%, and the storage requirements are reduced through the hash compression algorithm. Detailed implementation manners

[0045] The present invention will be further described in combination with specific embodiments.

[0046] A method for short video splicing and distribution, characterized by comprising the following steps:

[0047] Construct an intelligent hierarchical material library, and the intelligent hierarchical material library includes a material management system in six dimensions, namely frame picture materials, background video materials, dynamic patch materials, background music materials, transition effect materials and dynamic text materials;

[0048] Implement intelligent combination and duplication elimination of each material in the intelligent hierarchical material library through a dynamic combination duplication elimination engine. The dynamic combination duplication elimination engine adopts a dual-factor duplication elimination mechanism of a frequency control module and a content detection module to ensure the uniqueness of the intelligent combination of each material and the content presentation in the generated short video;

[0049] The intelligent distribution control system completes the efficient and compliant distribution of short video content through a multi-level state transition and real-time decision-making mechanism.

[0050] The six dimensions of the intelligent hierarchical material library include:

[0051] The first dimension of frame picture materials is subjected to RGB color gamut standardization processing with a resolution not lower than 1080P (1920×1080 pixels), the color gamut range is sRGB IEC61966-2.1, and an invisible digital watermark is embedded in combination with SHA-256 hash value verification. The embedding strength of the digital watermark in the DCT domain is ≥0.3;

[0052] The second dimension background video material adopts H.265 / HEVC encoding format, CRF value ≤23, key frame interval is controlled within 0.5 seconds, and copyright management is performed based on Hyperledger Fabric blockchain evidence storage technology. Each video is bound to an independent usage counter, and the threshold for each video within 24 hours is ≤3 times;

[0053] The third dimension dynamic patch material has a transparency dynamic adjustment range of 20%-80%, a step value of 5%, and a size that does not exceed 30% of the main screen area. It uses version 4 of the UUID unique identifier binding and device fingerprint dual verification. Device fingerprint dual verification refers to the IMEI+MAC address hash mixing;

[0054] The fourth dimension background music material has a fixed sampling rate of 44.1kHz, a bit rate of no less than 128kbps, and VBR mode is turned off. It also uses audio fingerprint recognition technology to achieve cross-platform authorization tracking. The peak feature extraction accuracy of the audio fingerprint recognition technology is ±0.01dB.

[0055] The fifth dimension transition special effects material has a special effects duration range of 0.2-1.5 seconds and an accuracy of 0.1s. It supports multi-layer Alpha channel superposition, but a maximum of 8 layers. It also uses discrete reorganization technology to prevent the reuse of special effects templates. The element replacement rate in the discrete reorganization technology is ≥ 65%;

[0056] The sixth dimension is dynamic text material. The font library is limited to the commercial authorized whitelist (such as Source Han Sans and Alibaba Puhui font). The text displacement path is parameterized (Bezier curve accuracy 0.1px), and the text content hash value (MD5) is matched with the sensitive word blacklist database in real time. The update frequency is ≤5 minutes.

[0057] The dynamic combination deduplication engine adopts a dual-factor deduplication mechanism of a frequency control module and a content detection module to ensure that the generated short video is unique in terms of material combination and content presentation, specifically:

[0058] The frequency control module dynamically tracks the number of times and time intervals of each material being called, ensuring that the reuse interval of elements in the same material group is ≥ the preset threshold, which is a generation cycle of 20 times; it uses a sliding window counting algorithm to count the usage of each material in the most recent N generations, where N is configurable and the default value = 50;

[0059] The content detection module, based on multi-modal feature extraction technology, performs hash encoding and similarity comparison on each material combination, including the extraction and comparison of visual features, audio features, and text features. Specifically, the visual features are the extraction of PHash (Perceptual Hash) and color histograms of frame pictures or videos, the audio features are the generation of Mel spectrogram fingerprints of background music, and the text features are the TF-IDF vectorization after word segmentation of dynamic text content;

[0060] The deduplication process logic includes steps of material preselection and frequency verification, combined hash generation, historical record comparison, content difference enhancement, and record update and locking;

[0061] Material preselection and frequency verification: According to the template definition, randomly select candidate materials from the corresponding material groups, such as background video materials, dynamic patch materials, etc.; query the real-time usage records of the candidate materials;

[0062] Combined hash generation: Generate a composite hash identifier for the currently selected material set;

[0063] Historical record comparison: Retrieve in the distributed hash database whether there are similar records. If there are matching records, it is determined as a potentially duplicate combination and enters the dynamic adjustment process;

[0064] Content difference enhancement: Apply random perturbation visual adjustment to frame picture materials, with a hue shift of ±5% and Gaussian noise σ = 0.01; perform layout variation on the position random offset of dynamic patch materials, with an X / Y axis offset of ±3% of the canvas size; perform timing perturbation on the transition effect duration with a floating range of ±0.2 seconds, and an audio track delay of ±100ms;

[0065] Record update and locking: Write the finally determined material combination hash value into the Redis cache with a TTL of 24 hours, and synchronize it to the blockchain node for evidence storage; and update the corresponding material usage counter;

[0066] The exception handling and degradation strategy includes candidate exhaustion handling, real-time performance guarantee, and threshold dynamic adjustment;

[0067] Candidate exhaustion handling: If the deduplication check fails continuously for 5 times, start the cross-group material call and select alternative materials from the associated material groups;

[0068] Real-time performance guarantee: Use the Bloom Filter to pre-screen historical hashes, reducing the database query time from an average of 12ms to 0.3ms;

[0069] Threshold dynamic adjustment: Automatically adjust the similarity threshold according to the platform algorithm feedback (such as duplicate rate warning), with an initial value of 0.15 and a floating range of 0.1 - 0.2;

[0070] The following technical effects are achieved through the dynamic combination and duplicate elimination engine: the average time consumption for single duplicate elimination check is reduced from 45 ms in the traditional solution to 8 ms; under the generation volume of tens of millions, the content duplication rate is stabilized below 3.2%, while it is 58 - 72% in the traditional solution; the memory occupancy is reduced by 42%, and the storage requirement is reduced through the hash compression algorithm.

[0071] The intelligent distribution control system includes four core processing stages and three auxiliary control modules:

[0072] The four core processing stages include the to-be-generated queue management, real-time copyright verification, priority distribution execution, and post-distribution processing;

[0073] The three auxiliary control modules include the exception rejection processing module, timeliness control module, and data feedback module;

[0074] The efficient and compliant distribution of short video content is realized through the four core processing stages and the three auxiliary control modules.

[0075] The to-be-generated queue management is as follows: the short videos generated by the dynamic duplicate elimination engine enter the to-be-distributed queue according to the preset priority rules. The default policy of the priority rules is: timeliness > copyright level > generation timestamp; the short video formats in the to-be-distributed queue are standardized and converted, and metadata (including generation time, template version, and copyright identification code) is injected, and the capacity of the distribution queue is controlled; the token bucket algorithm is used for capacity control to limit the length of the distribution queue, and when the threshold is exceeded, the automatic expansion or degradation strategy is triggered;

[0076] The real-time copyright verification is as follows: the authorization status of each material in the short video is queried through the blockchain node for material-level verification, and the response time ≤ 300 ms; the text content of the sensitive word library and the audio fingerprint for similarity detection with the copyright music library are compared in real time. The update period of the sensitive word library ≤ 5 minutes, and the similarity detection threshold < 0.1; the materials in the short video that pass the real-time copyright verification are marked as the "distributable" state, and the abnormal materials are transferred to the manual review queue;

[0077] The priority distribution execution adopts the dynamic scheduling algorithm, specifically: the basic weight is calculated by adding the weight coefficients of whether the user is a VIP, the linearly calculated value of the content validity period ratio, and the 24-hour distribution success rate. The weight coefficient of VIP is 0.4, the weight coefficient of non-VIP is 0, the weight coefficient of the linearly calculated value of the content validity period ratio is 0 - 0.3, and the weight coefficient of the 24-hour distribution success rate > 95% is 0.4, otherwise it is 0; the priority distribution is carried out according to the size of the basic weight value; the weight coefficients are dynamically updated every 5 minutes according to the network load and the platform interface status;

[0078] The post-distribution processing includes: for the materials in the successfully distributed short videos, the counter is immediately incremented to update the count. If the materials are for commercial use, they will be automatically marked as "disabled" after reaching the authorized upper limit of usage. For example, if the authorized upper limit is 3 times per day; a survival period is set for each material, and the survival period TTL = 1 hour. The materials that have not been distributed after the timeout are automatically transferred to cold storage and archived to different storage levels (SSD / HDD / offline tape library) according to the popularity grading.

[0079] The abnormal rejection processing module: for short videos with failed copyright verification, incompatible platform formats, and / or content compliance risks, automatically analyzes the reasons for rejection and generates repair suggestions. For items that can be automatically repaired, such as resolution adjustment, starts the transcoding service, which takes less than 8 seconds. After repair, the short video is re-added to the head of the distribution queue; items that require manual intervention are transferred to the review work order system and the operation personnel are notified.

[0080] The aging control module adopts a multi-level aging strategy, specifically: for urgent content, a countdown of 5 - 15 minutes is set, and if it has not been distributed after the timeout, a text message alarm is triggered; for regular content, a dynamic TTL mechanism is adopted, which is automatically adjusted according to the queue pressure for 1 - 6 hours; for long-tail content, a background silent distribution mode is enabled, and the bandwidth occupancy rate is less than 10%.

[0081] The data feedback module: monitors the distribution success rate (target value ≥ 99.5%), average response time (SLA standard < 800ms), and copyright dispute rate (threshold < 0.1%) in real time, adaptively optimizes the LSTM prediction model trained based on historical data, and gives an early warning of potential bottlenecks 10 minutes in advance; generates a distribution quality report every hour and automatically optimizes the priority algorithm parameters.

[0082] The interaction between the intelligent distribution control system and the dynamic combined deduplication engine receives the video hash value generated by the deduplication engine as the unique identifier of the distribution record; in real time, it feeds back the distribution success rate data to guide the dynamic combined deduplication engine to adjust the selection strategy of each material.

[0083] The linkage with copyright management: retrieves the latest authorized whitelist before distribution, and the synchronization period ≤ 1 minute. For each successfully distributed commercial material, an immutable record is written to the blockchain.

[0084] Abnormal state backtracking: establishes a full-link tracking ID, and can locate the specific link of distribution failure within 30 seconds.

[0085] Supports the historical record replay function, and the maximum backtracking period = 7 days.

[0086] The verification of the technical effects achieved by the present invention is as follows:

[0087] Distribution efficiency: The average response time is optimized from 3.2 seconds in the traditional solution to 720 milliseconds; the peak processing capacity of a single node reaches 1,200 requests per second;

[0088] Compliance guarantee: The copyright verification interception rate is increased from 65% in manual review to 99.8%; the incidence rate of copyright disputes after distribution is decreased to 0.07%;

[0089] System stability: The success rate remains >95% under 200% load pressure; the automatic scaling response time < 45 seconds.

[0090] A program for short video splicing and distribution, characterized by including computer-executable instructions, which, when executed by a processor, implement the method for short video splicing and distribution described above.

[0091] A storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the method for short video splicing and distribution described above.

Claims

1. A method for short video splicing and distribution, characterized in that, It includes the following steps: Construct an intelligent hierarchical material library, which contains a material management system with six dimensions, namely frame picture materials, background video materials, dynamic patch materials, background music materials, transition effect materials, and dynamic text materials; The intelligent combination and duplicate elimination of each material in the intelligent hierarchical material library are realized through a dynamic combination duplicate elimination engine. The dynamic combination duplicate elimination engine adopts a dual-factor duplicate elimination mechanism of a frequency control module and a content detection module to ensure the uniqueness of the intelligent combination and content presentation of the generated short videos; The intelligent distribution control system completes the efficient and compliant distribution of short video content through a multi-level state transfer and real-time decision-making mechanism.

2. The method for short video splicing and distribution according to claim 1, wherein The six dimensions of the intelligent hierarchical material library include: The first dimension of frame picture materials is processed by standardizing the RGB color gamut with a resolution of not less than 1080P, and an invisible digital watermark is embedded and combined with SHA-256 hash value verification; The second dimension of background video materials adopts the H.265 / HEVC encoding format, the key frame interval is controlled within 0.5 seconds, and the copyright management is carried out based on the blockchain deposit and proof technology of Hyperledger Fabric. Each video is bound with an independent usage counter; The third dimension of dynamic patch materials has a dynamic transparency adjustment range of 20%-80%, the size ratio does not exceed 30% of the main screen area, and the UUID unique identifier binding and device fingerprint double verification are adopted; The fourth dimension of background music materials has a fixed sampling rate of 44.1kHz and a bit rate of not less than 128kbps, and cross-platform authorization tracking is realized through audio fingerprint recognition technology; The fifth dimension of transition effect materials has an effect duration range of 0.2-1.5 seconds, supports multi-layer Alpha channel superposition, and adopts a discretization recombination technology to prevent the reuse of effect templates; The sixth dimension of dynamic text materials limits the font library to the commercial authorization whitelist, the text displacement path is parameterized, and the text content hash value is matched with the sensitive word blacklist database in real time.

3. The method for short video splicing and distribution according to claim 2, wherein The dynamic combination duplicate elimination engine adopts a dual-factor duplicate elimination mechanism of a frequency control module and a content detection module to ensure the uniqueness of the material combination and content presentation of the generated short videos. Specifically: The frequency control module dynamically tracks the call times and time intervals of each material to ensure that the reuse interval of elements in the same material group ≥ the preset threshold; the sliding window counting algorithm is used to count the usage of each material in the recent N generations; The content detection module, based on multi-modal feature extraction technology, performs hash encoding and similarity comparison on each material combination, including the extraction and comparison of visual features, audio features, and text features. The visual features are specifically the extraction of PHash and color histogram of frame pictures or videos, the audio features are specifically the generation of Mel spectrum fingerprints of background music, and the text features are specifically the TF-IDF vectorization after word segmentation of dynamic text content; The duplicate elimination process logic includes steps of material preselection and frequency verification, combined hash generation, historical record comparison, content difference enhancement, and record update and locking; The exception handling and degradation strategy include candidate exhaustion handling, real-time performance guarantee, and threshold dynamic adjustment.

4. The method for short video splicing and distribution according to claim 3, wherein, The intelligent distribution control system includes four core processing stages and three auxiliary control modules: The four core processing stages include to-be-generated queue management, real-time copyright verification, priority distribution execution, and post-distribution processing; The three auxiliary control modules include an exception rejection processing module, a timeliness control module, and a data feedback module; The efficient and compliant distribution of short video content is achieved through the four core processing stages and the three auxiliary control modules.

5. The method for short video splicing and distribution according to claim 4, wherein The to-be-generated queue management is as follows: Short videos generated by the dynamic duplicate removal engine enter the to-be-distributed queue according to the preset priority rules. The default policy of the priority rules is: timeliness > copyright level > generation timestamp. The video formats in the to-be-distributed queue are standardized and converted, metadata is injected, and the capacity of the distribution queue is controlled. The capacity control uses the token bucket algorithm to limit the length of the distribution queue. When the threshold is exceeded, an automatic expansion or downgrading strategy is triggered.

6. The method for short video splicing and distribution according to claim 4, characterized in that The real-time copyright verification is as follows: Query the authorization status of each material in the short video through the blockchain node for material-level verification; Compare the text content of the sensitive word library in real time and the audio fingerprint for similarity detection with the copyright music library; Materials in the short video that pass the real-time copyright verification are marked as the "distributable" state, and abnormal materials are transferred to the manual review queue.

7. The method for short video splicing and distribution according to claim 4, characterized in that, The priority distribution execution uses a dynamic scheduling algorithm, specifically: Calculate the basic weight by adding the weight coefficients of whether the user is a VIP, the linearly calculated value of the content validity period ratio, and the 24-hour distribution success rate. The weight coefficient of VIP is 0.4, and the weight coefficient of non-VIP is 0. The weight coefficient of the linearly calculated value of the content validity period ratio is 0 - 0.

3. The weight coefficient of the 24-hour distribution success rate > 95% is 0.4, otherwise it is 0; Perform priority distribution according to the magnitude of the basic weight value; Dynamically update the weight coefficients every 5 minutes according to the network load and the status of the platform interface.

8. The method for short video splicing and distribution according to claim 4, characterized in that, The post-distribution processing includes: Immediately trigger a counter increment operation for the materials in the successfully distributed short video to update the count. If the material is for commercial use, it is automatically marked as the "disabled" state after the cumulative usage reaches the authorization limit; Set the survival period for each material. Materials that are not distributed after the timeout are automatically transferred to cold storage and archived to different storage levels according to the popularity ranking.

9. The method for short video splicing and distribution according to claim 4, wherein The exception rejection processing module: For short videos with failed copyright verification, incompatible platform formats, and / or content compliance risks, automatically analyze the rejection reasons and generate repair suggestions. For items that can be automatically repaired, start the transcoding service, and after repair, add the short video back to the head of the to-be-distributed queue; Items that require manual intervention are transferred to the review work order system and the operation personnel are notified.

10. The method for short video splicing and distribution according to claim 4, characterized in that, The timeliness control module adopts a multi-level timeliness strategy, specifically: Set a countdown for urgent content, and trigger a text message alarm if it is not distributed after the timeout; For regular content, use the dynamic TTL mechanism and automatically adjust according to the queue pressure; For long-tail content, enable the background silent distribution mode.

11. The method for short video splicing and distribution according to claim 4, characterized in that, The data feedback module: Monitor the distribution success rate, average response time, and copyright dispute rate in real time, adaptively optimize the LSTM prediction model trained based on historical data, and give early warnings of potential bottlenecks; Generate a distribution quality report every hour and automatically optimize the parameters of the priority algorithm.

12. The method for short video splicing and distribution according to claim 1, wherein The dynamic combination deduplication engine uses a Bloom filter to pre-screen historical hashes, reducing the database query time from an average of 12 ms to 0.3 ms.

13. The method for short video splicing and distribution according to claim 1, wherein The intelligent distribution control system retrieves the latest authorized whitelist before distribution. Each time a commercial material is successfully distributed, an immutable record is written to the blockchain.

14. A program for short video splicing and distribution, characterized in that, It includes computer-executable instructions that, when executed by a processor, implement the method for short video splicing and distribution as described in any one of claims 1 to 13.

15. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for short video splicing and distribution as described in any one of claims 1 to 13.