Ultra-high-definition to high-definition multi-camera intelligent video generation method and system

Abstract

The invention discloses an ultrahigh-definition-to-high-definition multi-camera intelligent video generation method and system, and particularly relates to the technical field of video processing.The method comprises the steps that whether the video collection quality reaches the standard or not is judged based on the frame rate, resolution, noisy points and colors of collected ultrahigh-definition videos, and if yes, the collection quality evaluation coefficient of the videos is calculated; if the video acquisition quality does not reach the standard, executing different operations based on specific conditions until the video acquisition quality reaches the standard, converting an ultra-high-definition video into a high-definition video after determining that the video acquisition quality reaches the standard, calculating a peak signal-to-noise ratio and a structural similarity coefficient of a conversion link, comparing the peak signal-to-noise ratio and the structural similarity coefficient with a set standard, and judging whether the conversion effect reaches the standard or not; the image conversion reliability coefficient is calculated if the standard is reached, and the converted high-definition video is adjusted and optimized until the standard is reached if the standard is not reached, so that the video quality in the ultra-high-definition video acquisition and ultra-high-definition video conversion processes is improved to a certain extent, and the quality of the finally generated high-definition video is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of video processing, and more specifically, to a method and system for generating ultra-high definition to high definition multi-camera intelligent videos. Background Art

[0002] In the existing technical field of video processing, the conversion of ultra-high definition to high definition videos usually involves a simple downsampling process. However, this traditional method cannot effectively utilize the rich visual information in ultra-high definition videos to enhance the final high definition output. The intelligent video generation method for ultra-high definition to high definition has gradually become a research hotspot.

[0003] The existing methods for generating ultra-high definition to high definition multi-camera intelligent videos convert each frame of the captured ultra-high definition video into a high definition video frame image after capturing the ultra-high definition video, and then perform delay consistency correction on the generated high definition video frame images and store them in a database for subsequent production of high definition videos.

[0004] However, there are still some problems in the existing system: the existing system does not evaluate the quality of the captured ultra-high definition video after capturing the ultra-high definition video, and there may be a situation where the quality of the captured ultra-high definition videos is uneven, which will have a negative impact on the finally generated high definition video. In the process of converting ultra-high definition videos to high definition videos, the converted high definition videos are not analyzed either, and it cannot be guaranteed that the image quality of the converted high definition videos can be directly used for subsequent video generation, which may cause the quality of the finally generated high definition video to decline. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for generating ultra-high definition to high definition multi-camera intelligent videos to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for generating ultra-high definition to high definition multi-camera intelligent videos, comprising the following steps:

[0007] S1: Before shooting, measure and correct the on / off time deviation of each camera and the in / out time of the video image, and ensure that the on / off times of the multi-camera are consistent and the in / out times of the video recording images are consistent;

[0008] S2: Set the cameras of each position to turn on and off simultaneously at a specified time point of the GPS clock, and capture the ultra-high definition videos captured by the cameras of each position;

[0009] S3: Based on the frame rate, resolution, noise, and color of the ultra-high-definition videos captured by cameras at each position, determine whether the video capture quality meets the standard. If it meets the standard, calculate the video capture quality evaluation coefficient; if it does not meet the standard, perform different operations based on the specific situation until the video capture quality meets the standard;

[0010] S4: After determining that the video capture qualities all meet the standard, convert the ultra-high-definition videos into high-definition videos. After the conversion is completed, calculate the peak signal-to-noise ratio and structural similarity coefficient of the conversion process, and compare them with the set standards to determine whether the conversion effect meets the standard. If it meets the standard, calculate the image conversion reliability coefficient; if it does not meet the standard, adjust and optimize the converted high-definition videos until they meet the standard;

[0011] S5: After the conversion is completed, extract the material features and batch integrate the materials of the high-definition videos. Name the finally obtained basic material video frame dataset in the order of shooting date, shooting scene theme, and shooting material main name, and store it in the database. At the same time, calculate the material processing accuracy coefficient within the preset period;

[0012] S6: When making a video, retrieve the basic material video frame dataset from the database, then match and fuse the video frames and add effects. After generating the final high-definition video, back up the final high-definition video to the database;

[0013] S7: Collect the material processing accuracy coefficient within the preset period, the capture quality evaluation coefficient and total capture times of each ultra-high-definition video, and the image conversion reliability coefficient of each video conversion process;

[0014] S8: Calculate the average video capture quality evaluation coefficient, average video repeated capture rate coefficient, and average video image conversion reliability coefficient within the preset period, and further calculate the video generation quality maintenance index in combination with the material processing accuracy coefficient.

[0015] Preferably, the process of judging whether the video capture quality meets the standard in step S3 is as follows:

[0016] S31: Check the video frame rate and resolution of the j-th ultra-high-definition video captured, and compare them with the set video frame rate standard and resolution standard. If both are greater than or equal to the standards, the video frame rate and video resolution meet the standard; otherwise, re-capture the j-th ultra-high-definition video to determine whether the video frame rate and video resolution meet the standard;

[0017] S32: Use video analysis software to analyze the pixel changes in the video frames, quantify the number of pixel points mxdji and the number of noise points mzdji in the i-th video frame of the j-th ultra-high-definition video, and calculate the noise ratio αzbji in the i-th video frame of the j-th ultra-high-definition video. The specific formula is: Compare the calculated noise ratio in the $i$-th video frame of the $j$-th ultra-high-definition video with the set noise ratio standard in the $i$-th video frame of the $j$-th ultra-high-definition video. If the calculated value is greater than or equal to the preset value, the video frame contains excessive noise; otherwise, the video frame has qualified noise. Use color analysis software to analyze the color saturation, brightness, and hue in the video frame, and set the color evaluation standard according to the current industry standard for video production. If the color in the video frame meets the color evaluation standard, the color is qualified. When the video frame contains excessive noise or the color does not meet the standard, go to step S23;

[0018] S33. When the video frame contains excessive noise, mark the video frame with excessive noise and perform denoising processing on it until the noise ratio in the video frame is less than the corresponding set noise ratio standard. When the color of the video frame does not meet the standard, mark the video frame with non-compliant color and perform color correction on it until the color of the video frame meets the standard;

[0019] S34. When the video frame rate, resolution, noise, and color of the captured video all meet the standards, determine that the video capture quality meets the standards.

[0020] Preferably, when the video capture quality meets the standards, the specific formula for calculating the capture quality evaluation coefficient $R_{zbj}$ of the $j$-th ultra-high-definition video is: where $zlcj$, $zlej$, $fblcj$, $fblej$, and $nzj$ are the actual video frame rate, set video frame rate standard, actual video resolution, set video resolution, and number of video frames of the $j$-th ultra-high-definition video respectively, and $\alpha_{zej i}$, $\alpha_{scji}$ are the set noise ratio standard in the $i$-th video frame of the $j$-th ultra-high-definition video and the color compliance coefficient in the $i$-th video frame of the $j$-th ultra-high-definition video, with a value of 1.

[0021] Preferably, the specific conversion process in step S4 is as follows:

[0022] S41. Reduce the resolution of the ultra-high-definition video with qualified capture quality to the set standard video resolution at the high-definition level;

[0023] S42. Calculate the mean square error $MSE_{ji}$ between the $k$-th pixel point in the $i$-th video frame of the $j$-th ultra-high-definition video and the corresponding $k$-th pixel point in the $i$-th video frame of the $j$-th high-definition video after conversion. The specific formula is: $nzji$, $Ajik$, and $Bjik$ respectively refer to the number of pixel points contained in the $i$-th video frame of the $j$-th high-definition video, the pixel of the $k$-th pixel point in the $i$-th video frame of the $j$-th ultra-high-definition video, and the pixel of the $k$-th pixel point in the $i$-th video frame of the $j$-th high-definition video;

[0024] S43. Calculate the peak signal-to-noise ratio $\alpha_{fji}$ between the $i$-th video frame of the $j$-th ultra-high-definition video and the $i$-th video frame of the $j$-th high-definition video after conversion. The specific formula is: L represents the maximum numerical value of the pixel values of the pixels in the video frame;

[0025] S44. Calculate the structural similarity coefficient αsji between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video. The specific formula is as follows: μAji and μBji are the means of the i-th video frame image of the j-th ultra-high-definition video and the i-th video frame image of the j-th high-definition video respectively, and σAji 2 and σBji 2 are the variances of the i-th video frame image of the j-th ultra-high-definition video and the i-th video frame image of the j-th high-definition video respectively, σABji is the covariance between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video, and C1 and C2 are positive constants;

[0026] S45. Compare the calculated peak signal-to-noise ratio and structural similarity coefficient with the set lower limit values of the peak signal-to-noise ratio and the structural similarity coefficient respectively. If both calculated values are greater than the set values, the conversion effect meets the standard; otherwise, the conversion effect does not meet the standard.

[0027] S46. After the conversion effect meets the standard, calculate the image conversion reliability coefficient Rhj for converting the j-th ultra-high-definition video into a high-definition video. The specific formula is as follows:

[0028] Preferably, the specific steps for processing the high-definition video in step S5 are as follows:

[0029] S51. Automatically retrieve the target material name and extract all camera video frames containing the target material;

[0030] S52. Stitch the camera video frames containing the target material at the same time point to obtain a multi-camera video frame group of the scene picture containing the target material at the same time point;

[0031] S53. Delete the duplicate multi-camera video frame groups of the scene pictures containing the target material. Mark the multiple multi-camera video frame groups of the scene pictures containing the target material in the same video as the basic material video frame data set. The multi-camera video frame groups of the scene pictures containing the target material in the basic material video frame data set are numbered in sequence according to the time movement trajectory of the target material. The finally obtained basic material video frame data set is named and stored in the database in the order of shooting date, shooting scene theme, and shooting material main name;

[0032] S54, based on the theoretical number of video frames containing target material extracted in the i-th target material extraction process within the preset period, the actual number of video frames containing target material extracted, mtxi, the theoretical number of video frames spliced ​​in the j-th multi-camera video frame splicing process at the same time point, mbyj, the actual number of video frames spliced, mbxj, and the number of repetitions mcfk of the k-th multi-camera video frame group containing the target material, the material processing accuracy coefficient RCU is calculated. The specific formula is: nt, np, and nh are respectively the number of times the target material is extracted within a preset period, the number of times multi-camera video frames are spliced ​​at the same time point, and the number of multi-camera video frame groups of scene images containing the target material.

[0033] Preferably, the data processing process in step S8 is as follows:

[0034] S81, calculating the average video acquisition quality assessment coefficient Rzbs, the average video repetitive acquisition rate coefficient Rccs and the average video image conversion reliability coefficient Rhs within a preset period, the specific formula is as follows: nc is the number of videos collected in the preset period, and mzcj is the total number of times the jth video is collected;

[0035] S82. Calculate the video generation quality maintenance index QRS. The specific formula is:

[0036] To achieve the above object, the present invention provides the following technical solution: an ultra-high-definition to high-definition multi-camera intelligent video generation system, implementing the above ultra-high-definition to high-definition multi-camera intelligent video generation method, comprising:

[0037] Time consistency correction module: before shooting, the on / off time deviation of each camera and the video screen entry / exit time are measured and corrected to ensure that the on / off time of multiple cameras is consistent and the video recording screen entry / exit time is consistent;

[0038] Ultra-high-definition video acquisition module: used to acquire ultra-high-definition videos shot by cameras at various locations;

[0039] Video acquisition quality judgment module: judges whether the video acquisition quality meets the standards based on the frame rate, resolution, noise and color of the ultra-high-definition videos captured by the cameras at each position, and calculates the video acquisition quality evaluation coefficient if the video acquisition quality meets the standards;

[0040] Ultra-high-definition video conversion module: used to convert ultra-high-definition video into high-definition video. After the conversion is completed, the peak signal-to-noise ratio and structural similarity coefficient of the conversion link are calculated respectively, and compared with the set standard to determine whether the conversion effect meets the standard. If it meets the standard, the image conversion reliability coefficient is calculated. If it does not meet the standard, the converted high-definition video is adjusted and optimized until it meets the standard;

[0041] HD video conversion processing module: After the conversion is completed, extract the material features and batch integrate the materials of the HD video, name the finally obtained basic material video frame dataset in the order of shooting date, shooting scene theme, and shooting material main name, and store it in the database;

[0042] Material processing precision coefficient calculation module: Used to calculate the material processing precision coefficient within a preset period;

[0043] HD video generation data collection module: Used to collect the material processing precision coefficient within a preset period, the acquisition quality evaluation coefficient of each ultra-HD video, the total number of acquisitions, and the image conversion reliability coefficient of each video conversion link;

[0044] Video generation quality maintenance index calculation module: Used to calculate the average acquisition quality evaluation coefficient, the average repeated acquisition rate coefficient, and the average image conversion reliability coefficient of the video within a preset period, and further calculate the video generation quality maintenance index in combination with the material processing precision coefficient.

[0045] Technical effects and advantages of the present invention:

[0046] 1. Based on the frame rate, resolution, noise, and color of the ultra-HD videos captured by the cameras at each position, the present invention determines whether the video acquisition quality meets the standard. If it meets the standard, calculate the acquisition quality evaluation coefficient of the video; if it does not meet the standard, perform different operations based on the specific situation until the video acquisition quality meets the standard. After determining that the video acquisition quality all meets the standard, convert the ultra-HD video into an HD video, calculate the peak signal-to-noise ratio and structural similarity coefficient of the conversion link, and compare them with the set standard to determine whether the conversion effect meets the standard. If it meets the standard, calculate the image conversion reliability coefficient; if it does not meet the standard, adjust and optimize the converted HD video until it meets the standard. To a certain extent, it improves the quality of the captured ultra-HD videos and the video quality in the process of converting ultra-HD videos into HD videos, thereby improving the quality of the finally generated HD videos.

[0047] 2. After the conversion is completed, the present invention extracts the material features and batch integrates the materials of the HD video, names the finally obtained basic material video frame dataset in the order of shooting date, shooting scene theme, and shooting material main name, and stores it in the database. When generating videos subsequently, the target material video frame dataset can be quickly retrieved from the database, eliminating a large number of video frame deletion links and multi-camera video frame splicing links, reducing the video generation time, and improving the video generation efficiency. Description of the Drawings

[0048] Figure 1 It is a method step diagram of the present invention.

[0049] Figure 2 This is the system structure block diagram of the present invention. Detailed implementation manners

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0051] As Figure 1 shown, this embodiment provides a method for generating an ultra-high-definition to high-definition multi-camera intelligent video, including the following steps:

[0052] S1: Before shooting, measure and correct the on-off time deviation and video frame in-out time of each camera, and ensure that the on-off times of the multi-camera are consistent and the video recording frame in-out times are consistent.

[0053] Specifically, in this embodiment, the specific steps for measuring and correcting the on-off time deviation and video frame in-out time of each camera in step S1 are as follows:

[0054] S11: Randomly select several even-numbered time points under the GPS clock, and sequentially use them as the on-off adjustment points of each camera according to the time sequence. Control the switches of each camera according to the generated on-off time points, record the actual on-off time points of each camera in sequence, calculate the average on-off time deviation of each camera, mark it as the on-off time deviation of each camera, and correct the on-off times of each camera in sequence based on the marked information.

[0055] S12: After the on-off time correction of each camera is completed, randomly select several fixed-duration time periods under the GPS clock, set the same shooting object for each camera, switch each camera according to the start and end time points of the time period, record the video frame entry time and video frame disappearance time of each camera in each fixed-duration time period, subtract the video frame entry time from the on-time of each camera and subtract the video frame disappearance time from the off-time of each camera to obtain the video frame entry delay and video frame disappearance delay of each camera respectively. Calculate the average video frame entry delay and average video frame disappearance delay of each camera respectively, mark them as the video frame entry delay and video frame disappearance delay of each camera, and correct the video frame entry and disappearance times of each camera in sequence based on the marked information.

[0056] S2: Set the cameras at each position to turn on and off simultaneously at the specified time point of the GPS clock, and collect the ultra-high-definition videos captured by the cameras at each position.

[0057] S3: Based on the frame rate, resolution, noise, and color of the ultra-high-definition videos captured by the cameras at each position, determine whether the video capture quality meets the standard. If it meets the standard, calculate the acquisition quality evaluation coefficient of the video. If it does not meet the standard, perform different operations based on the specific situation until the video capture quality meets the standard.

[0058] Furthermore, the process of determining whether the video capture quality meets the standard in step S3 is as follows:

[0059] S31: Check the video frame rate and resolution of the j-th ultra-high-definition video captured, and compare them with the set video frame rate standard and resolution standard. If both are greater than or equal to the standard, the video frame rate and video resolution meet the standard. Otherwise, re-capture the j-th ultra-high-definition video to determine whether the video frame rate and video resolution meet the standard.

[0060] S32: Use video analysis software to analyze the pixel changes in the video frames, quantify the number of pixel points mxdji and the number of noise points mzdji in the i-th video frame of the j-th ultra-high-definition video, and calculate the noise ratio αzbji in the i-th video frame of the j-th ultra-high-definition video. The specific formula is: Compare the calculated noise ratio in the i-th video frame of the j-th ultra-high-definition video with the set noise ratio standard in the i-th video frame of the j-th ultra-high-definition video. If the calculated value is greater than or equal to the preset value, the video frame contains excessive noise. Otherwise, the video frame contains acceptable noise. Use color analysis software to analyze the color saturation, brightness, and hue in the video frame, and set the color evaluation standard based on the current industry standard for video production. If the color in the video frame meets the color evaluation standard, the color meets the standard. When the video frame contains excessive noise or the color does not meet the standard, go to step S23.

[0061] S33: When the video frame contains excessive noise, mark the video frame with excessive noise and perform denoising processing until the noise ratio in the video frame is less than the corresponding set noise ratio standard. When the color of the video frame does not meet the standard, mark the video frame with non-compliant color and perform color correction until the color of the video frame meets the standard.

[0062] S34: When the captured video frame rate, resolution, noise, and color all meet the standard, determine that the video capture quality meets the standard.

[0063] Furthermore, when the video capture quality meets the standard, the specific formula for calculating the acquisition quality evaluation coefficient Rzbj of the j-th ultra-high-definition video is: Among them, zlcj, zlej, fblcj, fblej, and nzj are respectively the actual video frame rate, the set video frame rate standard, the actual video resolution, the set video resolution, and the number of video frames of the j-th ultra-high-definition video. αzeji and αscji are respectively the standard of the noise proportion in the i-th video frame of the j-th ultra-high-definition video and the color compliance coefficient of the i-th video frame of the j-th ultra-high-definition video, and the value is 1.

[0064] S4: After determining that the video acquisition quality meets the standards, convert the ultra-high-definition video into a high-definition video, calculate the peak signal-to-noise ratio and the structural similarity coefficient in the conversion process, compare them with the set standards to determine whether the conversion effect meets the standards. If it meets the standards, calculate the image conversion reliability coefficient; if it does not meet the standards, adjust and optimize the converted high-definition video until it meets the standards.

[0065] Furthermore, the specific conversion process in step S4 is as follows:

[0066] S41: Reduce the resolution of the ultra-high-definition video with qualified acquisition quality to the set standard video resolution at the high-definition level.

[0067] S42: Calculate the mean square error MSEji between the k-th pixel point in the i-th video frame of the j-th ultra-high-definition video and the corresponding k-th pixel point in the i-th video frame of the j-th high-definition video after conversion. The specific formula is: nzji, Ajik, and Bjik respectively refer to the number of pixel points contained in the i-th video frame of the j-th high-definition video, the pixel of the k-th pixel point in the i-th video frame of the j-th ultra-high-definition video, and the pixel of the k-th pixel point in the i-th video frame of the j-th high-definition video.

[0068] S43: Calculate the peak signal-to-noise ratio αfji between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video after conversion. The specific formula is: L represents the maximum value of the pixel values of the pixel points in the video frame.

[0069] S44: Calculate the structural similarity coefficient αsji between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video. The specific formula is: μAji and μBji are respectively the mean values of the images of the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video, σAji 2 、σBji 2 are respectively the variances of the images of the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video, σABji is the covariance between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video, and C1 and C2 are positive constants.

[0070] Specifically, in this embodiment, the mean, variance, and covariance between two images can be directly obtained by existing algorithms. Now, they are regarded as known values and no specific value limits are set here.

[0071] S45. Compare the calculated peak signal-to-noise ratio and structural similarity coefficient with the set lower limit values of the peak signal-to-noise ratio and the structural similarity coefficient respectively. If both calculated values are greater than the set values, the conversion effect meets the standard; otherwise, the conversion effect does not meet the standard.

[0072] S46. After the conversion effect meets the standard, calculate the image conversion reliability coefficient Rhj for converting the j-th ultra-high-definition video into a high-definition video. The specific formula is as follows:

[0073] S5: After the conversion is completed, perform material feature extraction and material batch integration on the high-definition video. Name the finally obtained basic material video frame dataset in the order of shooting date, shooting scene theme, and shooting material main name and store it in the database. At the same time, calculate the material processing accuracy coefficient within a preset period.

[0074] Further, the specific steps for processing the high-definition video in step S5 are as follows:

[0075] S51. Automatically retrieve the target material name and extract all camera video frames containing the target material.

[0076] S52. Stitch the camera video frames containing the target material at the same time point to obtain a multi-camera video frame group of the scene picture containing the target material at the same time point.

[0077] S53. Delete the duplicate multi-camera video frame groups of the scene picture containing the target material. Mark the obtained multiple multi-camera video frame groups of the scene picture containing the target material in the same video as the basic material video frame dataset. The multi-camera video frame groups of the scene picture containing the target material in the basic material video frame dataset are numbered in sequence according to the time movement trajectory of the target material. The finally obtained basic material video frame dataset is named in the order of shooting date, shooting scene theme, and shooting material main name and stored in the database.

[0078] S54. Based on the theoretically extracted number of video frames mtyi containing the target material and the actually extracted number of video frames mtxi containing the target material during the i-th target material extraction process within a preset period, the theoretically stitched number of video frames mbyj and the actually stitched number of video frames mbxj during the j-th multi-camera video frame stitching process at the same time point, and the duplicate number mcfk of the k-th multi-camera video frame group of the scene picture containing the target material, calculate the material processing accuracy coefficient RCU. The specific formula is as follows: nt, np, and nh are respectively the number of times of target material extraction, the number of multi-camera video frame stitching at the same time point, and the number of multi-camera video frame groups of the scene picture containing the target material within a preset period.

[0079] S6: When making a video, after retrieving the basic material video frame dataset from the database, perform matching fusion and effect addition on the video frames, and after generating the final high-definition video, back up the final high-definition video to the database.

[0080] S7: Collect the material processing accuracy coefficient, the acquisition quality evaluation coefficient of each ultra-high-definition video, the total number of acquisitions, and the image conversion reliability coefficient of each video conversion link within a preset period.

[0081] S8: Calculate the average acquisition quality evaluation coefficient, the average repeated acquisition rate coefficient, and the average image conversion reliability coefficient of the video within a preset period respectively, and further calculate the video generation quality maintenance index in combination with the material processing accuracy coefficient.

[0082] Furthermore, the data processing process in step S8 is as follows:

[0083] S81: Calculate the average acquisition quality evaluation coefficient Rzbs, the average repeated acquisition rate coefficient Rccs, and the average image conversion reliability coefficient Rhs of the video within a preset period. The specific formulas are as follows: nc is the number of videos acquired within a preset period, and mzcj is the total number of acquisitions of the jth video acquired.

[0084] S82: Calculate the video generation quality maintenance index QRS. The specific formula is:

[0085] In this embodiment, it should be specifically noted that the set values and constant values used are selected based on actual needs, and no specific value limitations are made here.

[0086] Such as Figure 2The present embodiment provides a multi-camera intelligent video generation system for converting ultra-high definition to high definition, including a time consistency correction module, an ultra-high definition video acquisition module, a video acquisition quality judgment module, an ultra-high definition video conversion module, a converted high definition video processing module, a material processing precision coefficient calculation module, a high definition video generation data collection module, a video generation quality maintenance index calculation module, and a database. The time consistency correction module, the ultra-high definition video acquisition module, the video acquisition quality judgment module, the ultra-high definition video conversion module, the converted high definition video processing module, and the material processing precision coefficient calculation module are connected in sequence. The video acquisition quality judgment module, the ultra-high definition video conversion module, and the material processing precision coefficient calculation module are all connected to the high definition video generation data collection module. The high definition video generation data collection module is connected to the video generation quality maintenance index calculation module. All modules in the system are connected to the database.

[0087] The time consistency correction module measures and corrects the on / off time deviation of cameras at each shooting position and the in / out time of video frames before shooting, and determines that the on / off times of multi-camera are consistent and the in / out times of video recording frames are consistent.

[0088] The ultra-high definition video acquisition module is used to acquire ultra-high definition videos shot by cameras at each shooting position.

[0089] The video acquisition quality judgment module judges whether the video acquisition quality meets the standard based on the frame rate, resolution, noise, and color of the ultra-high definition videos shot by cameras at each shooting position, and calculates the acquisition quality evaluation coefficient of the video after the video acquisition quality meets the standard.

[0090] The ultra-high definition video conversion module is used to convert ultra-high definition videos into high definition videos. After the conversion is completed, the peak signal-to-noise ratio and structural similarity coefficient of the conversion link are calculated respectively, and compared with the set standard to judge whether the conversion effect meets the standard. If it meets the standard, the image conversion reliability coefficient is calculated. If it does not meet the standard, the converted high definition video is adjusted and optimized until it meets the standard.

[0091] After the conversion of the converted high definition video processing module is completed, it extracts the material features and batch integrates the materials of the high definition video, names the finally obtained basic material video frame data set in the order of shooting date, shooting scene theme, and shooting material main name, and stores it in the database.

[0092] The material processing precision coefficient calculation module is used to calculate the material processing precision coefficient within a preset period.

[0093] The high definition video generation data collection module is used to collect the material processing precision coefficient within a preset period, the acquisition quality evaluation coefficient of each ultra-high definition video and the total acquisition times, and the image conversion reliability coefficient of each video conversion link.

[0094] The video generation quality maintenance index calculation module is used to calculate the video average acquisition quality evaluation coefficient, the video average repeated acquisition rate coefficient, and the video average image conversion reliability coefficient within a preset period, and further calculate the video generation quality maintenance index in combination with the material processing accuracy coefficient;

[0095] The database is used to store the data information of all modules in the system.

[0096] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for generating ultra-high-definition to high-definition multi-camera intelligent videos, characterized in that: The following steps are involved: S1: Before shooting, measure and calibrate the on / off time deviation of each camera and the video screen entry / exit time to ensure that the on / off time of multiple cameras is consistent and the video recording screen entry / exit time is consistent; S2: Set the cameras at each position to be turned on and off at the same time at the specified time point of the GPS clock, and collect the ultra-high-definition videos shot by the cameras at each position; S3: judging whether the video acquisition quality meets the standards based on the frame rate, resolution, noise, and color of the ultra-high-definition videos captured by the cameras at each position; if the standards are met, calculating the video acquisition quality evaluation coefficient; if the standards are not met, performing different operations based on the specific situation until the video acquisition quality meets the standards; S4: After determining that the video acquisition quality meets the standard, convert the ultra-high-definition video into a high-definition video. After the conversion is completed, calculate the peak signal-to-noise ratio and structural similarity coefficient of the conversion link, and compare them with the set standard to determine whether the conversion effect meets the standard. If it meets the standard, calculate the image conversion reliability coefficient. If it does not meet the standard, adjust and optimize the converted high-definition video until it meets the standard; S5: After the conversion, the high-definition video is subjected to material feature extraction and material batch integration. The basic material video frame data set finally obtained is named in the order of shooting date, shooting scene theme and shooting material subject name and stored in the database. At the same time, the material processing accuracy coefficient within the preset period is calculated; S6: when making a video, a basic material video frame data set is retrieved from the database, and then matching and fusing the video frames and adding effects are performed. After the final high-definition video is generated, the final high-definition video is backed up in the database; S7: Collect the material processing accuracy coefficient within a preset period, the acquisition quality assessment coefficient of each ultra-high-definition video and the total number of acquisitions, and the image conversion reliability coefficient of each video conversion link; S8: Calculate the average video acquisition quality assessment coefficient, the average video repetition acquisition rate coefficient, and the average video image conversion reliability coefficient within a preset period respectively, and further calculate the video generation quality maintenance index in combination with the material processing accuracy coefficient.

2. The method for generating ultra-high-definition to high-definition multi-camera intelligent videos according to claim 1, characterized in that: The process of judging whether the video acquisition quality meets the standard in step S3 is as follows: S31, checking the video frame rate and resolution of the j-th ultra-high-definition video collected, and comparing them with the set video frame rate standard and resolution standard. If both are greater than or equal to the standard, the video frame rate and video resolution meet the standard. Otherwise, recollecting the j-th ultra-high-definition video determines whether the video frame rate and video resolution meet the standard. S32, using video analysis software to analyze pixel changes in video frames, quantify the number of pixels mxdji and the number of noise points mzdji in the i-th video frame of the j-th ultra-high-definition video, and calculate the noise point ratio αzbji in the i-th video frame of the j-th ultra-high-definition video. The specific formula is: The calculated noise percentage in the i-th video frame of the j-th ultra-high-definition video is compared with the set noise percentage standard in the i-th video frame of the j-th ultra-high-definition video. If the calculated value is greater than or equal to the preset value, the video frame contains excessive noise. Otherwise, the video frame contains noise that meets the standard. The color saturation, brightness and hue in the video frame are analyzed using color analysis software, and the color evaluation standard is set according to the current industry standard for video production. If the color in the video frame meets the color evaluation standard, the color meets the standard. When the video frame contains excessive noise or the color does not meet the standard, go to step S23. S33, when the video frame contains excessive noise, mark the video frame with excessive noise and perform denoising on it until the proportion of noise in the video frame is less than the corresponding set noise proportion standard; when the color of the video frame does not meet the standard, mark the video frame with insufficient color and perform color correction on it until the color of the video frame meets the standard; S34: When the frame rate, resolution, noise, and color of the captured video all meet the standards, it is determined that the video capture quality meets the standards.

3. The method for generating ultra-high-definition to high-definition multi-camera intelligent videos according to claim 2, characterized in that: When the video acquisition quality meets the standard, the specific formula for calculating the acquisition quality evaluation coefficient Rzbj of the jth ultra-high-definition video is: Wherein zlcj, zlej, fblcj, fblej and nzj are respectively the actual video frame rate, the set video frame rate standard, the actual video resolution, the set video resolution and the number of video frames of the j-th ultra-high-definition video; αzeji and αscji are respectively the set noise proportion standard in the ith video frame of the j-th ultra-high-definition video and the color compliance coefficient of the ith video frame of the j-th ultra-high-definition video, and their values ​​are 1.

4. The method for generating ultra-high-definition to high-definition multi-camera intelligent videos according to claim 3, characterized in that: The specific conversion process in step S4 is as follows: S41, reducing the resolution of ultra-high-definition video that meets the acquisition quality standards to a set standard video resolution of high-definition level; S42, calculating the mean square error MSEji between the kth pixel point in the i-th video frame of the j-th ultra-high-definition video and the corresponding kth pixel point in the i-th video frame of the j-th high-definition video after conversion, the specific formula is: nzji, Ajik and Bjik refer to the number of pixels contained in the ith video frame of the jth high-definition video, the kth pixel of the ith video frame of the jth ultra-high-definition video, and the kth pixel of the ith video frame of the jth high-definition video, respectively; S43, calculating the peak signal-to-noise ratio αfji between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video after conversion, the specific formula is: L represents the maximum pixel value of a pixel in a video frame; S44, calculating the structural similarity coefficient αsji between the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video, the specific formula is: μAji and μBji are the mean values ​​of the i-th video frame image of the j-th ultra-high-definition video and the i-th video frame image of the j-th high-definition video, respectively. σAji 2 ,σBji 2 are the variance of the i-th video frame image of the j-th ultra-high-definition video and the variance of the i-th video frame image of the j-th high-definition video, respectively. σABji is the covariance of the i-th video frame of the j-th ultra-high-definition video and the i-th video frame of the j-th high-definition video. C1 and C2 are positive constants. S45, respectively comparing the calculated peak signal-to-noise ratio and the structural similarity coefficient with the set peak signal-to-noise ratio lower limit and the structural similarity coefficient lower limit. If the calculated values ​​are both greater than the set values, the conversion effect meets the standard; otherwise, the conversion effect does not meet the standard. S46. After the conversion effect reaches the standard, the image conversion reliability coefficient Rhj of the j-th ultra-high-definition video converted into a high-definition video is calculated. The specific formula is:

5. The method for generating ultra-high-definition to high-definition multi-camera intelligent videos according to claim 4, characterized in that: The specific steps of processing the high-definition video in step S5 are as follows: S51, automatically searching for the target material name, and extracting all camera video frames containing the target material; S52, splicing the video frames of each camera position containing the target material at the same time point to obtain a multi-camera video frame group of the scene picture containing the target material at the same time point; S53, deleting duplicate scene picture multi-camera video frame groups containing target materials, marking multiple scene picture multi-camera video frame groups containing target materials in the same video as basic material video frame data sets, and numbering the scene picture multi-camera video frame groups containing target materials in the basic material video frame data sets in sequence according to the time motion trajectory of the target materials, and finally naming the basic material video frame data sets in the order of shooting date, shooting scene theme and shooting material subject name and storing them in the database; S54, based on the theoretical number of video frames containing target material extracted in the i-th target material extraction process within the preset period, the actual number of video frames containing target material extracted, mtxi, the theoretical number of video frames spliced ​​in the j-th multi-camera video frame splicing process at the same time point, mbyj, the actual number of video frames spliced, mbxj, and the number of repetitions mcfk of the k-th multi-camera video frame group containing the target material, the material processing accuracy coefficient RCU is calculated. The specific formula is: nt, np, and nh are respectively the number of times the target material is extracted within a preset period, the number of times multi-camera video frames are spliced ​​at the same time point, and the number of multi-camera video frame groups of scene images containing the target material.

6. The method for generating ultra-high-definition to high-definition multi-camera intelligent videos according to claim 5, characterized in that: The data processing process in step S8 is as follows: S81, calculating the average video acquisition quality assessment coefficient Rzbs, the average video repetitive acquisition rate coefficient Rccs and the average video image conversion reliability coefficient Rhs within a preset period, the specific formula is as follows: nc is the number of videos collected in the preset period, and mzcj is the total number of times the jth video is collected; S82. Calculate the video generation quality maintenance index QRS. The specific formula is:

7. An ultra-high definition to high definition multi-camera intelligent video generation system, implementing an ultra-high definition to high definition multi-camera intelligent video generation method according to any one of claims 1 to 6, characterized in that: include: Time consistency correction module: before shooting, the on / off time deviation of each camera and the video screen entry / exit time are measured and corrected to ensure that the on / off time of multiple cameras is consistent and the video recording screen entry / exit time is consistent; Ultra-high-definition video acquisition module: used to acquire ultra-high-definition videos shot by cameras at various locations; Video acquisition quality judgment module: judges whether the video acquisition quality meets the standards based on the frame rate, resolution, noise and color of the ultra-high-definition videos captured by the cameras at each position, and calculates the video acquisition quality evaluation coefficient if the video acquisition quality meets the standards; Ultra-high-definition video conversion module: used to convert ultra-high-definition video into high-definition video. After the conversion is completed, the peak signal-to-noise ratio and structural similarity coefficient of the conversion link are calculated respectively, and compared with the set standard to determine whether the conversion effect meets the standard. If it meets the standard, the image conversion reliability coefficient is calculated. If it does not meet the standard, the converted high-definition video is adjusted and optimized until it meets the standard; Conversion HD video processing module: After the conversion, the HD video is subjected to material feature extraction and batch integration, and the basic material video frame data set finally obtained is named in the order of shooting date, shooting scene theme and shooting material subject name and stored in the database; Material processing accuracy coefficient calculation module: used to calculate the material processing accuracy coefficient within a preset period; High-definition video generation data collection module: used to collect the material processing accuracy coefficient within a preset period, the acquisition quality assessment coefficient and total acquisition times of each ultra-high-definition video, and the image conversion reliability coefficient of each video conversion link; Video generation quality maintenance index calculation module: used to calculate the average video acquisition quality assessment coefficient, the average video repetition acquisition rate coefficient and the average video image conversion reliability coefficient within a preset period, and further calculate the video generation quality maintenance index in combination with the material processing accuracy coefficient.

Images