Short video clip synthesis method based on fuzzy logic

Abstract

The invention discloses a short video editing synthesis method based on fuzzy logic. The method comprises the following steps: S1, carrying out digital video preprocessing on an input original short video; s2, forming a short video feature data set; s3, generating a fuzzy logic decision result; s4, performing multi-target and multi-parameter optimization on the fuzzy logic decision result by adopting a random score programming method to form an optimized editing parameter set; s5, generating a preliminarily synthesized short video; s6, generating a video quality evaluation result; and S7, carrying out iterative optimization on the preliminarily synthesized short video according to a video quality evaluation result, adjusting the editing parameters and re-executing the editing synthesis step until the video quality evaluation result meets a preset standard, and generating a final edited finished product. According to the method, the editing scheme can achieve the best balance in the aspects of visual rhythm, information integrity and emotion expression consistency.

Description

Technical Field

[0001] The present invention relates to the technical field of short video editing, and in particular to a short video editing synthesis method based on fuzzy logic. Background Art

[0002] With the rapid development of the short video industry, the creation and optimization technology of short video content has received increasing attention. As a key link in short video production, the quality of short video editing directly affects the user's viewing experience, content dissemination effect and user stickiness of the platform. Existing short video editing methods mainly rely on manual editing or automated editing tools based on simple rules. However, there are many limitations when responding to large-scale short video production needs.

[0003] At present, manual editing is still the main way of creating short videos. Editors need to manually select editing points, adjust transition effects, and optimize audio and video synchronization based on the video content. Although manual editing can fully reflect the creator's subjective aesthetics and creative expression, its high time cost and labor intensity limit the mass production of short videos. In addition, there are large differences in experience and style among different editors, resulting in uneven editing quality, making it difficult to form a stable editing standard.

[0004] In order to solve the inefficiency of manual editing, some automated editing technologies have emerged, which automatically edit short videos based on fixed rules or simple machine learning models. For example, some automatic editing tools set editing points by detecting violent movements in the video, changes in audio rhythm or preset time nodes to achieve basic short video editing. However, the existing methods have a low level of intelligence and it is difficult to accurately grasp the emotional rhythm, content logic and visual beauty of the video. Especially when faced with complex video materials, existing automatic editing methods often cannot make intelligent judgments based on the video content, resulting in stiff and incoherent connections between edited segments, affecting the audience's viewing experience.

[0005] In addition, existing technologies also have significant deficiencies in video content understanding and optimization. Many automatic editing tools mainly rely on key frame extraction or simple audio peak analysis for editing and ignore the emotional information, color changes, and key rhythmic features in the video. For example, editing methods based on key frame detection will ignore important transitional scenes in the video, and audio peak analysis often finds it difficult to accurately match the content context of the video, resulting in a lack of coherence and narrative logic in the editing results. At the same time, existing technologies pay little attention to the audience's viewing psychology and the dissemination characteristics of short videos, and lack a comprehensive evaluation mechanism for the quality of video editing, resulting in uneven quality of the edited products.

[0006] In general, the existing short video editing technology has many limitations in terms of efficiency, intelligence and editing quality, and it is difficult to meet the needs of the modern short video industry for efficient, intelligent and high-quality editing. Therefore, a more intelligent editing method is urgently needed to better balance the content expression, visual rhythm and audience experience of short videos, and improve the automation level and quality control capabilities of editing. Summary of the invention

[0007] One object of the present invention is to propose a short video editing synthesis method based on fuzzy logic, which enables the editing scheme to achieve an optimal balance in terms of visual rhythm, information integrity and consistency of emotional expression.

[0008] A short video editing and synthesis method based on fuzzy logic of the present invention comprises the following steps: S1. Perform digital video preprocessing on the input original short video to extract frames, audio signals and subtitles; S2. Apply image processing and video analysis techniques to the short video preprocessing results to extract motion amplitude, visual features, color changes, audio rhythm and emotional information in the video frames to form a short video feature dataset; S3. Construct a fuzzy logic decision model based on the short video feature data set, use the preset membership function and fuzzy rules to quantify the extracted video features, determine the editing priority and transition requirements of each video clip, and generate the fuzzy logic decision results; S4. Use random fractional programming method to perform multi-objective and multi-parameter optimization on the fuzzy logic decision results, determine the best editing point, appropriate transition method and visual special effect application plan, and form an optimized editing parameter set; S5. Automatically edit and synthesize the short video preprocessing results according to the optimized editing parameter set, including the selection of editing points, the cropping of video clips, the application of transition effects and visual effects, and the synchronous synthesis of audio and video to generate a preliminary synthesized short video; S6. Conduct an overall quality assessment on the preliminary synthesized short video, evaluate it through a fuzzy logic evaluation system using video fluency, information integrity, and emotional expression consistency indicators, and generate a video quality assessment result; S7. Iteratively optimize the preliminary synthesized short video according to the video quality assessment result, adjust the editing parameters and re-execute the editing synthesis steps until the video quality assessment result meets the preset standard to generate the final editing product.

[0009] Optionally, the S1 includes the following steps: S11. Perform frame extraction processing on the input original short video, set the frame extraction interval to uniformly sample the video frames, and obtain the video frame sequence F; S12. Extract the audio signal from the original short video to obtain audio data A; S13. Extract subtitle information from short videos, parse the hidden subtitle track in the video, detect the text content in the video frame, and construct a subtitle dataset T.

[0010] Optionally, S2 includes the following steps: S21. Calculate the motion amplitude between adjacent frames based on the video frame sequence F : ; in, Indicates Frame image, Indicates Frame image, For Frame The set of pixels in Represents a pixel, and Represents pixel points The horizontal and vertical displacements between adjacent frames, and the calculated motion amplitude are used to characterize the dynamic characteristics of the video; S22. Video frame sequence Perform visual feature extraction and calculate the color histogram of each frame of image , which measures the change in visual features between adjacent frames: ; in, is the total number of color channels, and Represents frames and frame In the Normalized color histogram values ​​on color channels; S23. Calculate video frame sequence The color change characteristics of the frame are measured by the mean color offset: ; in, Representation frame In the The mean color value on the color channels, Representation frame In the The average color value on the color channel, the calculated color change characteristics Reflects the color adjustment information of the video frame; S24. Perform rhythm analysis on the audio signal A of the short video preprocessing result and calculate the energy envelope of the audio signal : ; in, For the Audio sampling point value; S25. Combined motion amplitude of video frames , Visual feature changes , color change characteristics and the audio signal energy envelope Construct a short video feature dataset: ; in, is the range of motion, For visual feature changes, For the color change characteristics, Audio energy.

[0011] Optionally, S3 includes the following steps: S31. Based on the short video feature dataset D, an improved hierarchical clustering algorithm is used to cluster the video features, and an improved fuzzy distance metric function is defined to calculate the fuzzy distance metric. : ; in, Represents the video feature vector, corresponding to the i-th motion amplitude , the i-th visual feature change , the i-th color change feature and the energy of the i-th audio , For the The weight of the feature, To be the first The fuzzy membership function of the item feature is mapped to the interval [0,1], is an adjustment factor used to balance the influence of Euclidean distance and fuzzy membership; S32. Using fuzzy distance measurement , the adaptive agglomerative hierarchical clustering algorithm is used to calculate the distance between any two clusters, and the distance between clusters is defined as: ; in, The clusters to be merged The clusters to be merged The inter-cluster distance, and They represent the clusters to be merged, For clustering The number of elements in Set the cluster merging threshold , when the distance between clusters When The clusters to be merged Merge until a set of clusters is generated: ; S33. For each cluster in the cluster set , calculate the mean of each video feature in the cluster and construct the cluster feature vector: ; in, is the average motion amplitude of the video clips in the i-th cluster, is the average visual feature change of the video clips in the i-th cluster, is the average color change feature of the video clips in the i-th cluster, is the average audio feature of the video clip in the i-th cluster; S34. Using each cluster feature vector Construct a cluster-level fuzzy logic decision model, and for each cluster Set the clustering feature vector as the fuzzy input variable and set the fuzzy output variable: ; in, Representation clustering The editing priority of the video clips in the Representation clustering The transition requirements of the internal video clips use a Gaussian fuzzy membership function for each input variable , at the same time, based on the preset improved fuzzy rules: ; in, Represent the fuzzy sets of each input variable, represents the corresponding output fuzzy set; S35. For each cluster The fuzzy logic decision results are weighted fusion method to obtain the global decision output and calculate the editing priority of the overall video clip. and transition requirements , generating the overall fuzzy logic decision result.

[0012] Optionally, the S35 includes the following steps: S351. Calculating weighted editing contribution of global editing priority based on hierarchical clustering structure and weighted transition demand contribution : ; ; in, For clustering The number of fragments in For clustering The average fuzzy distance from all fragments to the cluster center, For clustering The variance of the blur distances of all fragments in , and The regulating factors S352: Calculate global clip priority using weighted fusion method : ; in, The clip priority obtained by fuzzy logic reasoning; S353, using weighted fusion method to calculate global transition requirements : ; in, The transition requirements are obtained by fuzzy logic reasoning; S354. Calculate the adaptive correction factor for the global clipping decision : ; The revised editing priorities and transition requirements are: ; ; S355, combined with global clip priority and transition requirements , generate fuzzy logic decision results: .

[0013] Optionally, S4 includes the following steps: S41. Establish a random fractional programming optimization model for editing points, transition methods and visual effects based on fuzzy logic decision results: ; in, is the editing parameter vector to be optimized, including editing point selection, transition mode and visual effects application strategy, Represents the uncertainty factor of short videos, and are the number of positive evaluation indicators and negative impact indicators for editing optimization, and is the weight parameter, represents the positive influence function of clip optimization, represents the negative influence function of clip optimization; S42. Calculate the forward optimization index set , define the positive influence function of clip optimization: ; in, represents the editing priority obtained by fuzzy logic reasoning, Represents the transition requirements obtained by fuzzy logic reasoning, Represents the visual smoothness of the short video clip after editing, which is determined by the optimization parameters And the uncertainty of short videos Decide, is the fuzzy membership function of the overall rhythm smoothness of the video, which is used to constrain the editing rhythm so that the distribution of editing points conforms to natural visual habits. , , All are weights; S43. Calculate the negative impact indicator set , define the negative influence function of clip optimization: ; in, and They are the preset target editing priority and target transition requirements, Represents possible visual disturbances caused by editing point selection and transition optimization. is the fuzzy membership function of the editing rhythm disruption degree, All are weights; S44. Use random fractional programming optimization method to define constraints: ; ; in, and Set the minimum and maximum possible values ​​of the clipping parameters respectively, Fuzzy membership value representing the smoothness of the clip rhythm, setting the threshold Ensure that the rhythm of the optimized editing plan meets the requirements of visual comfort; S45. The Monte Carlo method is used to solve the random score optimization problem and the optimization objective based on sample mean approximation is defined as: ; in, is the number of random samplings, Representative The uncertain factor samples of round sampling are used to obtain the optimal editing parameters through optimization. , and finally form the optimized editing parameter set: ; in, Represents the number of optimized clipping parameters, Representative Optimize editing parameters.

[0014] Optionally, S5 includes the following steps: S51. Determine the best editing point in the short video based on the optimized editing parameter set, and select the best editing point in order according to the editing priority, video motion characteristics, scene changes and audio rhythm; S52. Divide the short video into multiple video clips according to the selected optimal editing points. The editing process takes into account the integrity, information carrying capacity and picture coherence of the video clips, so that each clip is independent in terms of visual, narrative and emotional expression, while avoiding information loss or content fragmentation due to improper selection of editing points; S53. Select a suitable transition method in combination with the transition requirement information in the optimized editing parameter set. The selection of the transition method is based on the visual characteristics, color consistency, motion trajectory and emotional changes of the adjacent video clips, so that the connection between different clips is natural and smooth. In the transition optimization process, a buffer transition method is used for video clips with high motion amplitude, and a gradual transition strategy is used for clips with strong color contrast. The weights of different types of transition methods are adjusted according to the overall style of the short video; S54. Apply visual effects to specific video clips according to the optimized editing parameter set. The selection of visual effects is based on the editing priority, picture style and content expression requirements of the video clips. In the process of visual effects optimization, emphasis effects are applied to high-priority clips to enhance the expression of the content, and soft effects are applied to low-priority clips; S55. After the video clips are cropped and the visual effects are optimized, the audio part of the short video is adjusted synchronously. The audio and video are synthesized synchronously to match the audio rhythm with the picture rhythm. During the synchronization process, the audio rhythm is analyzed and the alignment of the audio track and the video clip is adjusted in combination with the motion rhythm of the video clip. For short videos with a strong sense of rhythm, the editing points are adjusted to match the strong beat position of the rhythm. For videos with strong narrative, the audio fade-in and fade-out time is fine-tuned to make the audio and video transition more natural. S56. The preliminary synthesis of the short video is completed by combining the results of editing point selection, video clip cropping, transition optimization, visual special effects application and audio and video synchronization synthesis. During the synthesis process, frame-level detail adjustments are made to the optimized video clips.

[0015] The beneficial effects of the present invention are: (1) The present invention introduces a fuzzy logic decision model, combines multiple key features of the video, such as motion amplitude, visual features, color changes, audio rhythm and emotional information, to construct fuzzy logic inference rules, and uses an improved fuzzy membership function to quantify the video features, thereby intelligently deciding the editing priority and transition requirements of each video clip. Existing video editing methods usually rely on fixed rules or simple key frame detection, and have limitations in content understanding and intelligent decision-making. It is difficult to accurately judge the editing timing and transition method of video clips. Through fuzzy logic modeling, editing decisions can adapt to different types of video content, improve the intelligence level of video editing, and ensure that the editing results are more logical, smooth and visually consistent.

[0016] (2) After extracting the short video features, the present invention uses an improved adaptive hierarchical clustering algorithm to perform cluster analysis on the video clips. By defining a fuzzy distance measurement function, the similarity of the video clips in terms of motion, color, and audio multi-dimensional features can be accurately quantified. Existing editing optimization methods are mainly based on fixed rules or simple machine learning models, which make it difficult to balance the local coherence and global optimization effects of editing. Through fuzzy distance calculation and adaptive agglomerative hierarchical clustering, the clips can be reasonably grouped according to content similarity, ensuring that the selection of editing points is consistent with the overall rhythm and visual style of the video.

[0017] (3) In the editing optimization process, the present invention adopts a random score programming method to perform multi-objective and multi-parameter optimization on editing points, transition methods and visual special effects. By constructing positive and negative impact indicators, weighted optimization is performed using editing priority, transition requirements and visual fluency key factors, and the Monte Carlo method is combined to solve the optimal editing parameters so that the editing scheme can achieve the best balance in terms of visual rhythm, information integrity and consistency of emotional expression. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 The present invention provides a flowchart of a short video editing and synthesis method based on fuzzy logic. DETAILED DESCRIPTION

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0020] refer to Figure 1 , a short video editing synthesis method based on fuzzy logic, comprising the following steps: S1. Perform digital video preprocessing on the input original short video to extract frames, audio signals and subtitles; S2. Apply image processing and video analysis techniques to the short video preprocessing results to extract motion amplitude, visual features, color changes, audio rhythm and emotional information in the video frames to form a short video feature dataset; S3. Construct a fuzzy logic decision model based on the short video feature data set, use the preset membership function and fuzzy rules to quantify the extracted video features, determine the editing priority and transition requirements of each video clip, and generate the fuzzy logic decision results; S4. Use random fractional programming method to perform multi-objective and multi-parameter optimization on the fuzzy logic decision results, determine the best editing point, appropriate transition method and visual special effect application plan, and form an optimized editing parameter set; S5. Automatically edit and synthesize the short video preprocessing results according to the optimized editing parameter set, including the selection of editing points, the cropping of video clips, the application of transition effects and visual effects, and the synchronous synthesis of audio and video to generate a preliminary synthesized short video; S6. Conduct an overall quality assessment on the preliminary synthesized short video, evaluate it through a fuzzy logic evaluation system using video fluency, information integrity, and emotional expression consistency indicators, and generate a video quality assessment result; S7. Iteratively optimize the preliminary synthesized short video according to the video quality assessment result, adjust the editing parameters and re-execute the editing synthesis steps until the video quality assessment result meets the preset standard to generate the final editing product.

[0021] In this implementation, S1 includes the following steps: S11. Perform frame extraction processing on the input original short video, set the frame extraction interval to uniformly sample the video frames, and obtain the video frame sequence F; S12. Extract the audio signal from the original short video to obtain audio data A; S13. Extract subtitle information from short videos, parse the hidden subtitle track in the video, detect the text content in the video frame, and construct a subtitle dataset T.

[0022] In this implementation, S2 includes the following steps: S21. Calculate the motion amplitude between adjacent frames based on the video frame sequence F : ; in, Indicates Frame image, Indicates Frame image, For Frame The set of pixels in Represents a pixel, and Represents pixel points The horizontal and vertical displacements between adjacent frames, and the calculated motion amplitude are used to characterize the dynamic characteristics of the video; S22. Video frame sequence Perform visual feature extraction and calculate the color histogram of each frame of image , which measures the change in visual features between adjacent frames: ; in, is the total number of color channels, and Represents frames and frame In the Normalized color histogram values ​​on color channels; S23. Calculate video frame sequence The color change characteristics of the frame are measured by the mean color offset: ; in, Representation frame In the The mean color value on the color channels, Representation frame In the The average color value on the color channel, the calculated color change characteristics Reflects the color adjustment information of the video frame; S24. Perform rhythm analysis on the audio signal A of the short video preprocessing result and calculate the energy envelope of the audio signal : ; in, For the Audio sampling point value; S25. Combined motion amplitude of video frames , Visual feature changes , color change characteristics and the audio signal energy envelope Construct a short video feature dataset: ; in, is the range of motion, For visual feature changes, For the color change characteristics, Audio energy.

[0023] In this implementation, S3 includes the following steps: S31. Based on the short video feature dataset D, an improved hierarchical clustering algorithm is used to cluster the video features, and an improved fuzzy distance metric function is defined to calculate the fuzzy distance metric. : ; in, Represents the video feature vector, corresponding to the i-th motion amplitude , the i-th visual feature change , the i-th color change feature and the energy of the i-th audio , For the The weight of the feature, To be the first The fuzzy membership function of the item feature is mapped to the interval [0,1], is an adjustment factor used to balance the influence of Euclidean distance and fuzzy membership; S32. Using fuzzy distance measurement , the adaptive agglomerative hierarchical clustering algorithm is used to calculate the distance between any two clusters, and the distance between clusters is defined as: ; in, The clusters to be merged The clusters to be merged The inter-cluster distance, and They represent the clusters to be merged, For clustering The number of elements in Set the cluster merging threshold , when the distance between clusters When The clusters to be merged Merge until a set of clusters is generated: ; S33. For each cluster in the cluster set , calculate the mean of each video feature in the cluster and construct the cluster feature vector: ; in, is the average motion amplitude of the video clips in the i-th cluster, is the average visual feature change of the video clips in the i-th cluster, is the average color change feature of the video clips in the i-th cluster, is the average audio feature of the video clip in the i-th cluster; S34. Using each cluster feature vector Construct a cluster-level fuzzy logic decision model, and for each cluster Set the clustering feature vector as the fuzzy input variable and set the fuzzy output variable: ; in, Representation clustering The editing priority of the video clips in the Representation clustering The transition requirements of the internal video clips use a Gaussian fuzzy membership function for each input variable , at the same time, based on the preset improved fuzzy rules: ; in, Represent the fuzzy sets of each input variable, represents the corresponding output fuzzy set; S34 Rule The conditional part (IF condition): 1. Motion amplitude condition: If clustering The average range of motion within Belongs to fuzzy sets , that is, the overall motion amplitude of the video clip meets the specific blur range; 2. Visual feature conditions: If clustering Average visual feature variation within Belongs to fuzzy sets , that is, the overall visual change degree of the video clip conforms to the preset fuzzy classification; 3. Color change condition: If clustering Average color change characteristics within Belongs to fuzzy sets , that is, the color change trend of the video clip conforms to the set fuzzy feature range; 4. Audio feature conditions: If clustering Average audio characteristics within Belongs to fuzzy sets , that is, the audio characteristics of the video clip conform to a specific fuzzy category.

[0024] Conclusion (THEN Conclusion): 1. Clip priority decision: Then the cluster Clip priority Belongs to fuzzy sets , that is, comprehensively judging the importance of a video clip in editing based on its motion, visual, color and audio features; Transition demand decision: Then the cluster Transition requirements Belongs to fuzzy sets That is, according to the overall characteristics of the video clip, the transition type and connection method required for the clip in the editing are determined.

[0025] S35. For each cluster The fuzzy logic decision results are weighted fusion method to obtain the global decision output and calculate the editing priority of the overall video clip. and transition requirements , generating the overall fuzzy logic decision result.

[0026] In this implementation, S35 includes the following steps: S351. Calculating weighted editing contribution of global editing priority based on hierarchical clustering structure and weighted transition demand contribution : ; ; in, For clustering The number of fragments in For clustering The average fuzzy distance from all fragments to the cluster center, For clustering The variance of the blur distances of all fragments in , and The regulating factors S352: Calculate global clip priority using weighted fusion method : ; in, The clip priority obtained by fuzzy logic reasoning; S353, using weighted fusion method to calculate global transition requirements : ; in, The transition requirements are obtained by fuzzy logic reasoning; S354. Calculate the adaptive correction factor for the global clipping decision : ; The revised editing priorities and transition requirements are: ; ; S355, combined with global clip priority and transition requirements , generate fuzzy logic decision results: .

[0027] In this implementation, S4 includes the following steps: S41. Establish a random fractional programming optimization model for editing points, transition methods and visual effects based on fuzzy logic decision results: ; in, is the editing parameter vector to be optimized, including editing point selection, transition mode and visual effects application strategy, Represents the uncertainty factor of short videos, and are the number of positive evaluation indicators and negative impact indicators for editing optimization, and is the weight parameter, represents the positive influence function of clip optimization, represents the negative influence function of clip optimization; S42. Calculate the forward optimization index set , define the positive influence function of clip optimization: ; in, represents the editing priority obtained by fuzzy logic reasoning, Represents the transition requirements obtained by fuzzy logic reasoning, Represents the visual smoothness of the short video clip after editing, which is determined by the optimization parameters And the uncertainty of short videos Decide, is the fuzzy membership function of the overall rhythm smoothness of the video, which is used to constrain the editing rhythm so that the distribution of editing points conforms to natural visual habits. , , All are weights; S43. Calculate the negative impact indicator set , define the negative influence function of clip optimization: ; in, and They are the preset target editing priority and target transition requirements, Represents possible visual disturbances caused by editing point selection and transition optimization. is the fuzzy membership function of the editing rhythm disruption degree, All are weights; S44. Use random fractional programming optimization method to define constraints: ; ; in, and Set the minimum and maximum possible values ​​of the clipping parameters respectively, Fuzzy membership value representing the smoothness of the clip rhythm, setting the threshold Ensure that the rhythm of the optimized editing plan meets the requirements of visual comfort; S45. The Monte Carlo method is used to solve the random score optimization problem and the optimization objective based on sample mean approximation is defined as: ; in, is the number of random samplings, Representative The uncertain factor samples of round sampling are used to obtain the optimal editing parameters through optimization. , and finally form the optimized editing parameter set: ; in, Represents the number of optimized clipping parameters, Representative Optimize editing parameters.

[0028] In this implementation, S5 includes the following steps: S51. Determine the best editing point in the short video based on the optimized editing parameter set, and select the best editing point in order according to the editing priority, video motion characteristics, scene changes and audio rhythm; S52. Divide the short video into multiple video clips according to the selected optimal editing points. The editing process takes into account the integrity, information carrying capacity and picture coherence of the video clips, so that each clip is independent in terms of visual, narrative and emotional expression, while avoiding information loss or content fragmentation due to improper selection of editing points; S53. Select a suitable transition method in combination with the transition requirement information in the optimized editing parameter set. The selection of the transition method is based on the visual characteristics, color consistency, motion trajectory and emotional changes of the adjacent video clips, so that the connection between different clips is natural and smooth. In the transition optimization process, a buffer transition method is used for video clips with high motion amplitude, and a gradual transition strategy is used for clips with strong color contrast. The weights of different types of transition methods are adjusted according to the overall style of the short video; S54. Apply visual effects to specific video clips according to the optimized editing parameter set. The selection of visual effects is based on the editing priority, picture style and content expression requirements of the video clips. In the process of visual effects optimization, emphasis effects are applied to high-priority clips to enhance the expression of the content, and soft effects are applied to low-priority clips; S55. After the video clips are cropped and the visual effects are optimized, the audio part of the short video is adjusted synchronously. The audio and video are synthesized synchronously to match the audio rhythm with the picture rhythm. During the synchronization process, the audio rhythm is analyzed and the alignment of the audio track and the video clip is adjusted in combination with the motion rhythm of the video clip. For short videos with a strong sense of rhythm, the editing points are adjusted to match the strong beat position of the rhythm. For videos with strong narrative, the audio fade-in and fade-out time is fine-tuned to make the audio and video transition more natural. S56. The preliminary synthesis of the short video is completed by combining the results of editing point selection, video clip cropping, transition optimization, visual special effects application and audio and video synchronization synthesis. During the synthesis process, frame-level detail adjustments are made to the optimized video clips.

[0029] Example 1: On the morning of January 15, 2024, the chest pain center of a tertiary hospital in City A welcomed a patient who was admitted to the hospital due to sudden chest pain. After the patient entered the emergency green channel, the hospital immediately started a standardized acute chest pain treatment process. During the treatment process, the hospital medical team collected the patient's ECG monitoring data, coronary angiography video, CT images, echocardiogram, and simultaneously recorded the doctor's consultation and discussion process. Since this case is a typical case of acute myocardial infarction, the hospital hopes to organize the entire treatment process into a high-quality medical teaching video for training purposes.

[0030] However, in the past, when hospitals produced such teaching videos, editors usually needed to manually screen key clips from massive videos, adjust editing points, and determine appropriate transitions. This process usually took more than 14 hours. In addition, since manual editing relies on the subjective judgment of editors, there are large differences in the style and content quality of videos produced by different editors.

[0031] The method of the present invention first pre-processes all recorded videos. The hospital's medical imaging management system provides the following data: ECG monitoring video: 10 minutes, recording the ECG changes of the patient from admission to treatment; Coronary angiography video: 22 minutes, including contrast agent injection, coronary artery blockage location display and surgical process; Echocardiogram video: 8 minutes, showing the real-time movement of the patient's heart; CT angiography video: 15 minutes, providing 3D reconstruction of coronary arteries; Doctors’ consultation and discussion video: 30 minutes, recording the discussion process of experts from the cardiology, emergency and imaging departments.

[0032] The data totals about 85 minutes. First, frame extraction and audio and video separation are performed to automatically identify key frames in medical images, medical terms in audio, and subtitle information for subsequent editing.

[0033] In video preprocessing, the fuzzy logic decision model analyzes the motion amplitude, color changes, and audio rhythm characteristics of all video clips and automatically selects the most critical clips. For example: In the ECG monitoring video, the key frames of ST segment elevation are automatically marked. The system detects that the ST segment is continuously elevated by more than 0.2mV and automatically selects the 15 seconds before and after this period as the editing segment; In the coronary angiography surgery video, the system analyzes the flow trajectory of the contrast agent and automatically selects the key 5 minutes of the surgery as a segment when it detects the diagnosis result of "99% stenosis of the proximal right coronary artery"; In the doctor consultation video, the system uses speech-to-text analysis technology to identify keywords such as "acute myocardial infarction", "PCI surgery", and "stent implantation", and automatically marks the time period of the discussion content, so that key discussions are retained; In the end, the system automatically edited a teaching video with a total length of 18 minutes, covering the entire treatment process from admission to the end of treatment.

[0034] In traditional manual editing, the transitions between different clips are usually selected manually by the editor based on experience, which may lead to abrupt transitions between some segments and affect the viewing experience. Based on the adaptive hierarchical clustering algorithm, the natural connection of the editing points is ensured, for example: Transition between ECG monitoring clips and coronary angiography clips: The "gradual fade-in" method is used to make the screen transition smoothly from the ECG to the angiography image, enhancing the viewing experience; Transition between doctor consultation video and imaging data: When doctors discuss key medical images, the system automatically inserts a picture-in-picture mode of the image on the right, allowing viewers to see the discussion and imaging data at the same time; After the editing is completed, the system conducts a quality assessment of the synthesized video to ensure that the video fluency, information completeness and visual coherence meet medical teaching standards.

[0035] Fluency score: Analyze the frame changes before and after the editing point during video playback. The fluency score is 93.4%, which is 12.7% higher than the existing method. Information completeness score: Medical experts reviewed and scored the information, with a score of 96.1%, which is 18.5% higher than the existing editing method; Viewing experience rating: 5 cardiology experts and 10 residents were invited to watch and rate the video, with an average score of 9.3 / 10, an improvement of 21.3% over existing methods. The comparative data analysis between the present invention and the traditional method is shown in the following table: Evaluation Items Traditional methods Method of the present invention Improvement effect Data sorting time 14 hours 2.8 hours Save 80.0% Editing Fluency Rating 82.9% 93.4% Increased by 12.7% Information completeness score 78.5% 96.1% Increase by 18.5% Viewing experience rating 7.6 points 9.3 points Increased by 21.3% Important fragment retention rate 72.1% 94.6% Increase by 22.5% The final video was uploaded to the hospital's internal medical teaching platform for young doctors to learn and use. It was also played as a typical case teaching video at the National Chest Pain Center Training Conference on February 20, and was highly recognized by experts and medical staff. Compared with traditional methods, automated editing technology not only greatly improves video production efficiency, but also ensures the integrity of medical information and viewing experience, demonstrating its wide application potential in the field of medical short video editing.

[0036] The present invention introduces a fuzzy logic decision model, combines multiple key features of the video, such as motion amplitude, visual features, color changes, audio rhythm and emotional information, to construct fuzzy logic inference rules, and uses an improved fuzzy membership function to quantify the video features, thereby intelligently deciding the editing priority and transition requirements of each video clip. Existing video editing methods usually rely on fixed rules or simple key frame detection, and have limitations in content understanding and intelligent decision-making. It is difficult to accurately judge the editing timing and transition method of video clips. Through fuzzy logic modeling, editing decisions can adapt to different types of video content, improve the intelligence level of video editing, and ensure that the editing results are more logical, smooth and visually consistent.

[0037] After extracting the short video features, the present invention uses an improved adaptive hierarchical clustering algorithm to perform cluster analysis on the video clips, and defines a fuzzy distance metric function to accurately quantify the similarity of the video clips in terms of motion, color, and audio multi-dimensional features. Existing editing optimization methods are mainly based on fixed rules or simple machine learning models, which make it difficult to balance the local coherence and global optimization effects of editing. Through fuzzy distance calculation and adaptive agglomerative hierarchical clustering, the clips can be reasonably grouped according to the similarity of the content, ensuring that the selection of editing points is consistent with the overall rhythm and visual style of the video.

[0038] In the editing optimization process, the present invention adopts a random score programming method to perform multi-objective and multi-parameter optimization on editing points, transition methods and visual special effects applications. By constructing positive and negative impact indicators, weighted optimization is performed using editing priorities, transition requirements, and key factors of visual fluency. The Monte Carlo method is combined to solve the optimal editing parameters so that the editing solution can achieve the best balance in terms of visual rhythm, information integrity and consistency of emotional expression.

[0039] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A short video editing and synthesis method based on fuzzy logic, characterized in that: The steps include: S1. Perform digital video preprocessing on the input original short video; S2. Apply image processing and video analysis techniques to the preprocessed original short video to form a short video feature dataset; S3. Construct a fuzzy logic decision model based on the short video feature data set, use the preset membership function and fuzzy rules to quantify the extracted video features, determine the editing priority and transition requirements of each video clip, and generate the fuzzy logic decision results; S4. Use random fractional programming method to perform multi-objective and multi-parameter optimization on the fuzzy logic decision results to form an optimized editing parameter set; S5. Automatically edit and synthesize the short video preprocessing results according to the optimized editing parameter set to generate a preliminary synthesized short video; S6. Conduct an overall quality assessment on the preliminary synthesized short video, evaluate it through a fuzzy logic evaluation system using video fluency, information integrity, and emotional expression consistency indicators, and generate a video quality assessment result; S7. Iteratively optimize the preliminary synthesized short video according to the video quality assessment result, adjust the editing parameters and re-execute the editing synthesis steps until the video quality assessment result meets the preset standard to generate the final editing product.

2. A short video editing and synthesis method based on fuzzy logic according to claim 1, characterized in that: The S1 comprises the following steps: S11. Perform frame extraction processing on the input original short video, set the frame extraction interval to uniformly sample the video frames, and obtain the video frame sequence F; S12. Extract the audio signal from the original short video to obtain audio data A; S13. Extract subtitle information from short videos, parse the hidden subtitle track in the video, detect the text content in the video frame, and construct a subtitle dataset T.

3. The short video editing and synthesis method based on fuzzy logic according to claim 2, characterized in that: The S2 comprises the following steps: S21. Calculate the motion amplitude between adjacent frames based on the video frame sequence F : ； in, Indicates Frame image, Indicates Frame image, For Frame The set of pixels in Represents a pixel, and Represents pixel points The horizontal and vertical displacements between adjacent frames, and the calculated motion amplitude are used to characterize the dynamic characteristics of the video; S22. Video frame sequence Perform visual feature extraction and calculate the color histogram of each frame image , which measures the change in visual features between adjacent frames: ； in, is the total number of color channels, and Represents frames and frame In the Normalized color histogram values ​​on color channels; S23. Calculate video frame sequence The color change characteristics of the frame are measured by the mean color offset: ； in, Representation frame In the The mean color value on the color channels, Representation frame In the The average color value on the color channel, the calculated color change characteristics Reflects the color adjustment information of the video frame; S24. Perform rhythm analysis on the audio signal A of the short video preprocessing result and calculate the energy envelope of the audio signal : ； in, For the Audio sampling point value; S25. Combined motion amplitude of video frames , Visual feature changes , color change characteristics and the audio signal energy envelope Construct a short video feature dataset: ； in, is the range of motion, For visual feature changes, For the color change characteristics, Audio energy.

4. The short video editing and synthesis method based on fuzzy logic according to claim 1, characterized in that: The S3 comprises the following steps: S31. Based on the short video feature dataset D, an improved hierarchical clustering algorithm is used to cluster the video features, and an improved fuzzy distance metric function is defined to calculate the fuzzy distance metric. : ； in, Represents the video feature vector, corresponding to the i-th motion amplitude , the i-th visual feature change , the i-th color change feature and the energy of the i-th audio , For the The weight of the feature, To be the first The fuzzy membership function that maps the item features to the interval [0,1], is an adjustment factor used to balance the influence of Euclidean distance and fuzzy membership; S32. Using fuzzy distance measurement , the adaptive agglomerative hierarchical clustering algorithm is used to calculate the distance between any two clusters, and the distance between clusters is defined as: ； in, The clusters to be merged The clusters to be merged The inter-cluster distance, and They represent the clusters to be merged, For clustering The number of elements in Set the cluster merging threshold , when the distance between clusters When The clusters to be merged Merge until a set of clusters is generated: ； S33. For each cluster in the cluster set , calculate the mean of each video feature in the cluster and construct the cluster feature vector: ； in, is the average motion amplitude of the video clips in the i-th cluster, is the average visual feature change of the video clips in the i-th cluster, is the average color change feature of the video clips in the i-th cluster, is the average audio feature of the video clip in the i-th cluster; S34. Using each cluster feature vector Construct a cluster-level fuzzy logic decision model, and for each cluster Set the clustering feature vector as the fuzzy input variable and set the fuzzy output variable: ； in, Representation clustering The editing priority of the video clips in the Representation clustering The transition requirements of the internal video clips use a Gaussian fuzzy membership function for each input variable , at the same time, based on the preset improved fuzzy rules: ； in, Represent the fuzzy sets of each input variable, represents the corresponding output fuzzy set; S35. For each cluster The fuzzy logic decision results are weighted fusion method to obtain the global decision output and calculate the editing priority of the overall video clip. and transition requirements , generating the overall fuzzy logic decision result.

5. The short video editing and synthesis method based on fuzzy logic according to claim 4, characterized in that: The S35 comprises the following steps: S351. Calculating weighted editing contribution of global editing priority based on hierarchical clustering structure and weighted transition demand contribution : ； ； in, For clustering The number of fragments in For clustering The average fuzzy distance from all fragments to the cluster center, For clustering The variance of the blur distances of all fragments in , and The regulating factors S352: Calculate global clip priority using weighted fusion method : ； in, The clip priority obtained by fuzzy logic reasoning; S353, using weighted fusion method to calculate global transition requirements : ； in, The transition requirements are obtained by fuzzy logic reasoning; S354. Calculate the adaptive correction factor for the global clipping decision : ； The revised editing priorities and transition requirements are: ； ； S355, combined with global clip priority and transition requirements , generate fuzzy logic decision results: 。 6. The short video editing and synthesis method based on fuzzy logic according to claim 1, characterized in that: The S4 comprises the following steps: S41. Establish a random fractional programming optimization model for editing points, transition methods and visual effects based on fuzzy logic decision results: ； in, Represents the weighted average of random variables, which is expected to measure the overall performance of editing decisions under different uncertainty conditions. is the editing parameter vector to be optimized, including editing point selection, transition mode and visual effects application strategy, Represents the uncertainty factor of short videos. and are the number of positive evaluation indicators and negative impact indicators for editing optimization, and is the weight parameter, represents the positive influence function of clip optimization, represents the negative influence function of clip optimization; S42. Calculate the forward optimization index set , define the positive influence function of clip optimization: ； in, represents the editing priority obtained by fuzzy logic reasoning, Represents the transition requirements obtained by fuzzy logic reasoning, Represents the visual smoothness of the short video clip after editing, which is determined by the optimization parameters And the uncertainty of short videos Decide, is the fuzzy membership function of the overall rhythm smoothness of the video, which is used to constrain the editing rhythm so that the distribution of editing points conforms to natural visual habits. , , All are weights; S43. Calculate the negative impact indicator set , define the negative influence function of clip optimization: ； in, and They are the preset target editing priority and target transition requirements, Represents possible visual disturbances caused by editing point selection and transition optimization. is the fuzzy membership function of the editing rhythm disruption degree, All are weights; S44. Use random fractional programming optimization method to define constraints: ； ； in, and Set the minimum and maximum possible values ​​of the clipping parameters respectively, Fuzzy membership value representing the smoothness of the clip rhythm, setting the threshold Ensure that the rhythm of the optimized editing plan meets the requirements of visual comfort; S45. The Monte Carlo method is used to solve the random score optimization problem and the optimization objective based on sample mean approximation is defined as: ； in, is the number of random samplings, Representative The uncertain factor samples of round sampling are used to obtain the optimal editing parameters through optimization. , and finally form the optimized editing parameter set: ； in, Represents the number of optimized clipping parameters, Representative Optimize editing parameters.

7. The short video editing and synthesis method based on fuzzy logic according to claim 1, characterized in that: The S5 comprises the following steps: S51. Determine the best editing point in the short video based on the optimized editing parameter set, and select the best editing point in order according to the editing priority, video motion characteristics, scene changes and audio rhythm; S52. Divide the short video into multiple video clips according to the selected optimal editing points. The editing process takes into account the integrity, information carrying capacity and picture coherence of the video clips, so that each clip is independent in terms of visual, narrative and emotional expression, while avoiding information loss or content fragmentation due to improper selection of editing points; S53. Select a suitable transition method in combination with the transition requirement information in the optimized editing parameter set. The selection of the transition method is based on the visual characteristics, color consistency, motion trajectory and emotional changes of the adjacent video clips, so that the connection between different clips is natural and smooth. In the transition optimization process, a buffer transition method is used for video clips with high motion amplitude, and a gradual transition strategy is used for clips with strong color contrast. The weights of different types of transition methods are adjusted according to the overall style of the short video; S54. Apply visual effects to specific video clips according to the optimized editing parameter set. The selection of visual effects is based on the editing priority, picture style and content expression requirements of the video clips. In the process of visual effects optimization, emphasis effects are applied to high-priority clips to enhance the expression of the content, and soft effects are applied to low-priority clips; S55. After the video clips are cropped and the visual effects are optimized, the audio part of the short video is adjusted synchronously. The audio and video are synthesized synchronously to match the audio rhythm with the picture rhythm. During the synchronization process, the audio rhythm is analyzed and the alignment of the audio track and the video clip is adjusted in combination with the motion rhythm of the video clip. For short videos with a strong sense of rhythm, the editing points are adjusted to match the strong beat position of the rhythm. For videos with strong narrative, the audio fade-in and fade-out time is fine-tuned to make the audio and video transition more natural. S56. The preliminary synthesis of the short video is completed by combining the results of editing point selection, video clip cropping, transition optimization, visual special effects application and audio and video synchronization synthesis. During the synthesis process, frame-level detail adjustments are made to the optimized video clips.

Images