Video storage optimization method and device, electronic equipment and storage medium

Abstract

The invention discloses a video storage optimization method and device, electronic equipment and a storage medium, and relates to the technical field of video storage. According to the application, environment context data such as illumination and weather can be combined, semantic analysis is performed on a real-time video by means of a multi-modal value perception model, and a value score is generated; and then a differentiated coding strategy and a classified storage mode are adopted according to the score, and extensive processing of unified compression and time-driven management is avoided, so that the problems of sharp storage cost increase, high management complexity, insufficient real-time performance and high reliability caused by the existing unified compression standard and time-driven management can be solved. According to the technical scheme of the invention, the technical problems of low storage cost, simplified management process, improved real-time performance of the system, guaranteed high-value video data quality and high-value density and storage efficiency dual demands are solved, and the technical effects of reducing the storage cost, simplifying the management process, improving the real-time performance of the system, and satisfying the high-value density and storage efficiency dual demands are achieved.

Description

Technical Field

[0001] This disclosure relates to the field of video storage technology, and in particular to a video storage optimization method and apparatus, electronic device and storage medium. Background Technology

[0002] Video surveillance systems, as a crucial infrastructure for security and smart cities, are widely used in traffic management, public safety, and industrial monitoring. Among related technologies, those employing unified compression standards and time-driven lifecycle management, while initially meeting basic storage needs through H.265 / HEVC encoding technology in the early stages of 4K / 8K ultra-high-definition camera adoption, have struggled to meet the dual demands of high-value density and storage efficiency as video data volumes have grown exponentially. This has led to industry challenges such as soaring storage costs, high management complexity, and insufficient system real-time performance. Summary of the Invention

[0003] This disclosure provides a method, apparatus, electronic device, and storage medium for optimizing video storage. Its main purpose is to address the industry challenges of soaring storage costs, high management complexity, and insufficient system real-time performance.

[0004] According to a first aspect of this disclosure, a method for optimizing video storage is provided, comprising: Acquire real-time video data and environmental context data, wherein the environmental context data at least includes information on light intensity, weather conditions, and scene sensitivity; Semantic analysis is performed on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data; wherein, the semantic analysis includes spatial feature extraction, temporal behavior analysis, and environmental context perception; Based on the value score, the real-time video data is subjected to hierarchical encoding of video regions; wherein different value scores correspond to different encoding strategies; The encoded real-time video data is classified and stored based on the value score; different value scores correspond to different storage categories.

[0005] Optionally, the step of performing semantic analysis on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data includes: A lightweight convolutional neural network model is used to extract spatial features from video frames of the real-time video data, identify specific targets in the video frames in real time, and label the semantic weights of the specific targets. The motion trajectory pattern of the specific target in the real-time video data is captured by a long short-term memory network, the behavioral risk value corresponding to the motion trajectory pattern is calculated, and a behavioral risk prediction value is generated based on a prediction algorithm. By combining the light intensity, weather conditions, and scene sensitivity information in the environmental context data, scene weighting coefficients are generated; The value score is calculated based on the specific target semantic weight, the behavioral risk value, and the scenario weight coefficient.

[0006] Optionally, performing hierarchical encoding of video regions on the real-time video data based on the value score includes: When the corresponding video region is determined to be a high-value video region based on the value score, lossless encoding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a medium-value video region based on the value score, medium compression coding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a low-value video region based on the value score, the Open Video Coding Standard is used for encoding.

[0007] Optionally, the step of classifying and storing the encoded real-time video data based on the value score includes: When a video region is determined to be a high-value video region based on the value score, the video data is stored in the edge node and the high-speed cloud storage medium; wherein, the retention period is set to permanent retention. When the corresponding video area is determined to be a medium-value video area based on the value score, the video data is stored in the local area network additional storage device. When the corresponding video region is determined to be a low-value video region based on the value score, the video data is stored in the edge storage node.

[0008] Optionally, the method further includes: A data cleanup task queue is created in the cache database, and expired video data is cleaned up based on the value score and the preset storage period strategy according to the preset period; wherein, the preset storage period is different for video data with different value scores.

[0009] According to a second aspect of this disclosure, a video storage optimization apparatus is provided, comprising: The acquisition unit is used to acquire real-time video data and environmental context data, wherein the environmental context data at least covers light intensity, weather conditions and scene sensitivity information; The analysis unit is used to perform semantic analysis on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data; wherein, the semantic analysis includes spatial feature extraction, temporal behavior analysis, and environmental context perception. The encoding unit is used to perform hierarchical encoding of video regions on the real-time video data according to the value score; wherein different value scores correspond to different encoding strategies; A storage unit is used to classify and store the encoded real-time video data based on the value score; wherein different value scores correspond to different storage categories.

[0010] Optionally, the analysis unit is further configured to: A lightweight convolutional neural network model is used to extract spatial features from video frames of the real-time video data, identify specific targets in the video frames in real time, and label the semantic weights of the specific targets. The motion trajectory pattern of the specific target in the real-time video data is captured by a long short-term memory network, the behavioral risk value corresponding to the motion trajectory pattern is calculated, and a behavioral risk prediction value is generated based on a prediction algorithm. By combining the light intensity, weather conditions, and scene sensitivity information in the environmental context data, scene weighting coefficients are generated; The value score is calculated based on the specific target semantic weight, the behavioral risk value, and the scenario weight coefficient.

[0011] Optionally, the encoding unit is further configured to: When the corresponding video region is determined to be a high-value video region based on the value score, lossless encoding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a medium-value video region based on the value score, medium compression coding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a low-value video region based on the value score, the Open Video Coding Standard is used for encoding.

[0012] Optionally, the storage unit is further used for: When a video region is determined to be a high-value video region based on the value score, the video data is stored in the edge node and the high-speed cloud storage medium; wherein, the retention period is set to permanent retention. When the corresponding video area is determined to be a medium-value video area based on the value score, the video data is stored in the local area network additional storage device. When the corresponding video region is determined to be a low-value video region based on the value score, the video data is stored in the edge storage node.

[0013] Optionally, the device further includes: The cleaning unit is used to create a data cleaning task queue in the cache database and perform expired video data cleaning based on the value score and the preset storage period strategy according to a preset period; wherein, the preset storage period is different for video data with different value scores.

[0014] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0015] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0016] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0017] The video storage optimization method, apparatus, electronic device, and storage medium disclosed herein, through this application, can combine environmental context data such as illumination and weather, perform semantic analysis on real-time video using a multimodal value perception model to generate value scores, and then adopt differentiated coding strategies and classified storage methods according to the scores, avoiding the coarse processing of uniform compression and time-driven management. Therefore, it can solve the technical problems of soaring storage costs, high management complexity, and insufficient real-time performance caused by existing uniform compression standards and time-driven management, making it difficult to meet the dual requirements of high value density and storage efficiency. It achieves the technical effects of reducing storage costs, simplifying management processes, improving system real-time performance, and ensuring the quality of high-value video data, thus meeting the dual requirements of high value density and storage efficiency.

[0018] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0019] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A schematic flowchart illustrating a video storage optimization method provided in an embodiment of this disclosure; Figure 2 A schematic diagram of the structure of a video storage optimization device provided in an embodiment of this disclosure; Figure 3 A schematic diagram of the structure of another video storage optimization device provided in an embodiment of this disclosure; Figure 4 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0020] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0021] The following description, with reference to the accompanying drawings, outlines a video storage optimization method, apparatus, electronic device, and storage medium according to embodiments of the present disclosure.

[0022] Figure 1 This is a schematic flowchart illustrating a video storage optimization method provided in an embodiment of the present disclosure.

[0023] like Figure 1 As shown, the method includes the following steps: Step 101: Obtain real-time video data and environmental context data, wherein the environmental context data at least covers light intensity, weather conditions and scene sensitivity information; The first step is to acquire real-time video data, which can be achieved through deployed surveillance cameras. These cameras can include 4K or 8K ultra-high-definition cameras, capable of capturing various scenes in real time. This includes the movement trajectories and details of dynamic targets such as people and vehicles, as well as image information of static scenes such as building walls and ground facilities. This provides continuous and complete raw data support for subsequent in-depth processing of the video content. Environmental context data is acquired, including light intensity data, which can be collected through the camera's built-in light sensor or external professional light detection equipment. This data accurately reflects the lighting conditions of the monitored scene, such as strong light during the day, weak light at dusk, and dark light at night. This data directly affects the brightness and detail clarity of the video image. Weather status data can be obtained by connecting to an external meteorological monitoring system or the environmental perception module integrated into the camera. This data can effectively identify various weather types, such as sunny, rainy, foggy, and snowy days. Different weather conditions will cause different interferences to the video image, such as rain obscuring the image or fog making the image blurry. Scene sensitivity information needs to be determined based on the preset attributes of the monitored scene. The scene can be divided into sensitive areas and ordinary areas. Sensitive areas include cash counter areas in financial institutions and core data center areas in enterprises. Ordinary areas include non-entrance corridors in residential areas and non-critical sections of urban roads. Scene information with different sensitivities provides an important environmental dimension basis for subsequent differentiated processing of video data.

[0024] Step 102: Perform semantic analysis on the real-time video data based on the multimodal value perception model to generate a value score for the real-time video data; wherein, the semantic analysis includes spatial feature extraction, temporal behavior analysis, and environmental context perception. The model uses a multimodal value perception model to perform semantic analysis on the acquired real-time video data. This model achieves a deep understanding of the video content by integrating multi-dimensional perception technology, rather than a simple analysis of a single dimension. It can more comprehensively mine the effective information in the video data to support subsequent value assessment.

[0025] The semantic analysis process specifically encompasses three parts: spatial feature extraction, temporal behavior analysis, and environmental context awareness. Spatial feature extraction employs a lightweight CNN model to identify specific targets in video frames in real time, such as faces, vehicles, and unusual objects. It assigns weights based on the semantic importance of different targets, ensuring accurate capture of key information in the video's spatial dimension. Temporal behavior analysis utilizes an LSTM network to continuously track the movement trajectories of targets in the video, identifying movement patterns such as loitering, running, and gathering, and then calculating corresponding behavioral risk values. This effectively perceives dynamic changes and potential risks in target behavior. Environmental context awareness combines previously acquired environmental context data with information such as light intensity, weather conditions, and scene sensitivity to generate scene weight coefficients that reflect the impact of the current environment on the video content.

[0026] After completing the above semantic analysis, the multimodal value perception model integrates the target semantic weights extracted from spatial features, the behavioral risk values ​​derived from temporal behavior analysis, and the scene weight coefficients generated by environmental context perception. It calculates and generates a value score for real-time video data through dynamically adjusted weight parameters. Furthermore, the weight parameters can be optimized in real time through reinforcement learning algorithms to adapt to the value assessment needs in different scenarios.

[0027] Step 103: Perform hierarchical encoding of video regions on the real-time video data based on the value score; wherein different value scores correspond to different encoding strategies; The core of performing hierarchical coding of video regions based on value scores is to match corresponding coding strategies for each region according to the semantic value differences of different regions in the real-time video data reflected by the value scores. This ensures the integrity of key information while minimizing data redundancy in non-critical regions and improving the overall utilization efficiency of storage resources.

[0028] When the value score is high, the corresponding video area usually contains key semantic information that is important for monitoring needs, such as faces and license plates. The encoding strategy should focus on preserving details and adopt processing methods that can accurately restore the information in the area, avoiding the loss of key information during the encoding process. When the value score is in the medium range, the corresponding video area has relatively moderate semantic importance. The encoding strategy needs to find a balance between image quality and data compression efficiency, ensuring that the image has the clarity required for monitoring while controlling the amount of data through appropriate compression to reduce unnecessary storage. When the value score is low, the corresponding video area is mostly static backgrounds and other content with low semantic value. The encoding strategy can focus on high compression rate processing, reducing storage consumption by significantly compressing the amount of data, and this high compression processing will not affect the overall monitoring function.

[0029] The entire layered encoding process is based on value scoring to ensure that video areas of different values ​​can receive appropriate encoding processing, thereby achieving synergistic optimization of storage efficiency and monitoring needs.

[0030] Step 104: Based on the value score, the encoded real-time video data is classified and stored; wherein, different value scores correspond to different storage categories.

[0031] The encoded real-time video data is categorized and stored based on value scoring. The core principle is to match different storage categories to the encoded video data according to the semantic importance reflected in the value score. This achieves precise allocation of storage resources, avoiding insufficient storage of high-value data or excessive resource consumption of low-value data, thus ensuring the economy and availability of surveillance data storage. When the value score is in the high range, the corresponding encoded video data usually contains critical monitoring information. Its storage category should prioritize security and longevity, selecting storage locations that provide a stable storage environment and support long-term retention, along with appropriate backup mechanisms to ensure data is not lost due to unforeseen circumstances.

[0032] When the value score is in the medium range, the semantic importance of the video data is moderate. The storage category can be chosen to balance storage costs and data retention requirements, setting a reasonable retention period to meet potential subsequent query and backtracking needs while avoiding excessive storage resource consumption. When the value score is in the low range, the video data is mostly content with low semantic value. The storage category can adopt lightweight processing, setting a shorter retention period, or even using overwrite storage to maximize storage space while ensuring basic monitoring data recording functions. The entire classification and storage process is based on the value score, ensuring that each type of encoded real-time video data corresponds to the appropriate storage category, achieving efficient utilization of storage resources.

[0033] In some embodiments, the step of performing semantic analysis on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data includes: A lightweight convolutional neural network model is used to extract spatial features from video frames of the real-time video data, identify specific targets in the video frames in real time, and label the semantic weights of the specific targets. The motion trajectory pattern of the specific target in the real-time video data is captured by a long short-term memory network, the behavioral risk value corresponding to the motion trajectory pattern is calculated, and a behavioral risk prediction value is generated based on a prediction algorithm. By combining the light intensity, weather conditions, and scene sensitivity information in the environmental context data, scene weighting coefficients are generated; The value score is calculated based on the specific target semantic weight, the behavioral risk value, and the scenario weight coefficient.

[0034] A lightweight convolutional neural network model is used to extract spatial features from video frames of real-time video data. This type of model has efficient real-time computing capabilities, can quickly process image information in video frames, and accurately identify specific targets within the frame. Common targets in monitoring scenarios, such as faces, vehicles, and abnormal objects, are all within the recognition range. At the same time, based on the importance of different targets to the monitoring needs, a corresponding semantic weight is labeled for each identified specific target, so that the semantic value of the target is clearly reflected.

[0035] By processing real-time video data through a Long Short-Term Memory (LSTM) network, the network can effectively capture information changes in the temporal dimension, continuously track the motion trajectory of a specific target in consecutive video frames, and then analyze different motion trajectory patterns such as loitering, running, and gathering. Based on the characteristics of these patterns, the network calculates the corresponding behavioral risk values. At the same time, it uses prediction algorithms to predict the target's subsequent possible behaviors, generates behavioral risk prediction values, and further improves the value assessment of the target's behavior.

[0036] By comprehensively analyzing environmental context data such as light intensity, weather conditions, and scene sensitivity information—for example, the impact of different light intensities on target recognition clarity, the interference of different weather conditions on image quality, and the differences in monitoring priority corresponding to different scene sensitivities—scene weight coefficients are generated based on these analysis results to reflect the impact of environmental factors on the value of video data. Finally, the semantic weights of specific targets, behavioral risk values, and scene weight coefficients are integrated, and these parameters are processed through a preset calculation method to obtain a value score for the real-time video data, ensuring that the score comprehensively and accurately reflects the semantic value of the video data.

[0037] In some embodiments, performing hierarchical encoding of video regions on the real-time video data based on the value score includes: When the corresponding video region is determined to be a high-value video region based on the value score, lossless encoding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a medium-value video region based on the value score, medium compression coding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a low-value video region based on the value score, the Open Video Coding Standard is used for encoding.

[0038] When performing hierarchical coding of video regions based on value scores, it is necessary to first determine the value level of the video region according to the value scores, and then match the corresponding coding method for different levels to ensure that the coding process is adapted to the value of the region.

[0039] When a video area is determined to be a high-value video area based on its value score, lossless encoding is performed using an efficient video encoding standard. The efficient video encoding standard has excellent detail restoration capabilities and can completely preserve key information within the high-value area, such as facial features and vehicle license plates in the monitoring scene. This information is crucial for subsequent monitoring backtracking and event investigation. Lossless encoding can prevent key details from being lost or blurred during the encoding process, ensuring the accuracy and integrity of the information.

[0040] When a video area is determined to be a medium-value video area based on its value score, a medium-compression encoding is performed using an efficient video encoding standard. This encoding method achieves a balance between picture quality and data compression efficiency. It ensures that the encoded video picture maintains the clarity required for daily monitoring and viewing, without affecting the identification of routine dynamics within the area. At the same time, it reduces the amount of data through appropriate compression, avoiding unnecessary storage resource occupation and adapting to the use and storage needs of medium-value areas.

[0041] When a video area is determined to be a low-value video area based on its value score, the Open Video Coding Standard (ODS) is used for encoding. The ODS has a high compression rate, which can significantly reduce the data size. Low-value video areas are mostly static backgrounds and other content with low semantic value. Even after high compression, it will not affect the overall monitoring function. At the same time, it can significantly reduce storage consumption and free up more resources for storing high- and medium-value video areas.

[0042] In some embodiments, classifying and storing the encoded real-time video data based on the value score includes: When a video region is determined to be a high-value video region based on the value score, the video data is stored in the edge node and the high-speed cloud storage medium; wherein, the retention period is set to permanent retention. When the corresponding video area is determined to be a medium-value video area based on the value score, the video data is stored in the local area network additional storage device. When the corresponding video region is determined to be a low-value video region based on the value score, the video data is stored in the edge storage node.

[0043] When classifying and storing encoded real-time video data based on value scoring, the storage scheme must be strictly matched to the value level of the video areas defined by the value score. When a video area is determined to be a high-value video area based on the value score, since the video data in this area usually contains key monitoring information such as faces and license plates, it is necessary to balance fast access and long-term security. Therefore, it is stored in edge nodes and high-speed cloud storage media. Edge nodes can achieve fast and local data retrieval to meet real-time monitoring needs, while high-speed cloud storage media can provide a stable and reliable long-term storage environment. At the same time, to ensure that key data is not lost, the retention period is set to permanent retention to ensure that subsequent query and backtracking needs can be met at any time.

[0044] When a video area is determined to be of medium value based on its value score, the semantic importance of the video data in this area is moderate. It does not need to be permanently retained but must meet query requirements within a certain period. The regional network's supplementary storage devices, with their centralized storage and easy-to-share access features, are well-suited to the storage needs of this type of data and can efficiently support the management and retrieval of daily monitoring data. When a video area is determined to be of low value based on its value score, this area mainly consists of static backgrounds and other content with low semantic value. Its storage priority is relatively low. Edge storage nodes, located close to the data acquisition end, have lower storage costs and can quickly complete data storage and subsequent overwrite updates, meeting the storage requirements of low-value video data and effectively avoiding the occupation of core storage resources.

[0045] In some embodiments, the method further includes: A data cleanup task queue is created in the cache database, and expired video data is cleaned up based on the value score and the preset storage period strategy according to the preset period; wherein, the preset storage period is different for video data with different value scores.

[0046] A data cleanup task queue is created in the cache database. The cache database is used to temporarily store encoded real-time video data, providing a convenient data access environment for cleanup operations and avoiding interference with routine business operations such as data reading and storage caused by direct operation of core storage devices. When creating the data cleanup task queue, key information of the video data to be cleaned, such as the unique data identifier, corresponding value score, and storage start time, must be entered into the queue in an orderly manner to ensure that cleanup tasks proceed in sequence and to prevent duplicate or missed cleanup.

[0047] Expired video data is cleaned up based on a preset cycle. The preset cycle can be set according to the operating load of the monitoring system. For example, it can be performed once every 12 hours or during the early morning when the system has less business. Performing cleanup during low-load periods can reduce the resource consumption of core business such as video data storage and retrieval, and ensure the stable operation of the system.

[0048] Different value ratings of video data correspond to different preset storage periods. This is because video data of different values ​​has different usage requirements in monitoring scenarios. High-value video data needs to be retained for a long time for subsequent queries and backtracking, so the preset storage period is longer; medium-value video data has a moderate retention period, which is sufficient to meet business needs within a certain period; low-value video data does not need to be stored for a long time, so the preset storage period is shorter. During cleanup, the system reads the value rating and storage period of each video data in the cache database and compares it with its corresponding preset storage period. If the data storage period exceeds the preset period, it is determined to be expired data, and then the expired data is deleted from the cache database and its corresponding storage location, achieving efficient reclamation of storage resources.

[0049] The following example illustrates the video storage optimization method disclosed in the embodiments of this application.

[0050] 1. AI Value Perception Module This module employs multimodal fusion analysis technology to achieve in-depth understanding and value quantification of video content: Spatial Feature Extraction: A lightweight CNN model (such as MobileNetV3) is used to identify specific targets (e.g., faces, vehicles, abnormal objects) in video frames in real time and to label their semantic weights. This technology draws inspiration from the video frame analysis technique mentioned in a patent from Anhui Tongshuo Information Technology Co., Ltd., but employs a more advanced model.

[0051] Temporal Behavior Analysis: LSTM networks are used to capture motion trajectory patterns (such as loitering, running, and gathering) and calculate behavioral risk values.

[0052] Contextual awareness: Generate scene weight coefficients by taking into account factors such as time (day / night), location (sensitive area / normal area), and weather conditions.

[0053] Value quantification formula: S_e = α·(object value) + β·(behavioral risk) + γ·(scenario weight) α, β, and γ are dynamic weight parameters that are optimized in real time through reinforcement learning algorithms to ensure that they accurately reflect the value of the content in different scenarios.

[0054] 2. Dynamic Encoding Strategy Engine Based on the value score S_e, this module implements an adaptive compression strategy.

[0055] 2.1 ROI Layered Coding Technology High-value areas (S_e ≥ 0.7): such as key information like faces and license plates, use H.266 / VVC lossless encoding (QP≤18) to ensure complete preservation of details.

[0056] Medium value region (0.3 ≤ S_e<0.7): Use H.265 / HEVC medium compression (QP=25-30) to balance quality and efficiency.

[0057] Low-value areas (S_e<0.3): such as static backgrounds, use AV1 or custom high-compression encoding (QP≥40) to significantly reduce the amount of data.

[0058] 2.2 Dynamic Frame Rate Control Technology High-activity scenarios (such as crowds): Maintain 30fps to ensure smoothness.

[0059] Low-activity scenarios (such as empty corridors): Reduce to 1-5fps to reduce frame redundancy.

[0060] This technology references the dynamic frame rate concept in the patent of Chengdu Anmusen Intelligent Technology, but adds semantic-based activity judgment.

[0061] 2.3 Environment-Adaptive Coding In low-light scenes: Activate the DeblurGAN deblurring algorithm to compensate for image quality rather than simply increasing the bitrate.

[0062] Rain and fog scenes: The DeRain algorithm is used to remove rain and fog while maintaining image clarity.

[0063] Ensure that the PSNR remains stable at ≥35dB under various environments.

[0064] Intelligent Storage Management System This system adopts a value-driven tiered storage strategy: Intelligent cleanup mechanism: Integrate a task scheduling system similar to OSS_Scheduler, create a cleanup task queue in Redis, and periodically clean up expired data according to a strategy to avoid duplicate operations.

[0065] Edge-Cloud Collaborative Processing Framework To improve system efficiency, a layered computing architecture is adopted: On-device (camera): Built-in lightweight AI chip (such as HiSilicon Hi3559A) performs real-time object detection and preliminary value scoring, reducing the burden on the backend.

[0066] Edge server: Receives data from the receiving end, performs complex multimodal AI analysis, generates a final value score, and performs hierarchical encoding and storage scheduling.

[0067] Cloud (Central Cloud): Receives high-value data for permanent storage, in-depth analysis (such as cross-camera tracking), and model training and updates.

[0068] Corresponding to the video storage optimization method described above, this invention also proposes a video storage optimization apparatus. Since the apparatus embodiments of this invention correspond to the method embodiments described above, details not disclosed in the apparatus embodiments can be referred to in the method embodiments described above, and will not be repeated here.

[0069] Figure 2 This is a schematic diagram of the structure of a video storage optimization device provided in an embodiment of the present disclosure, as shown below. Figure 2 As shown, it includes: Acquisition unit 21 is used to acquire real-time video data and environmental context data, wherein the environmental context data at least covers light intensity, weather conditions and scene sensitivity information; Analysis unit 22 is used to perform semantic analysis on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data; wherein, the semantic analysis includes spatial feature extraction, temporal behavior analysis and environmental context perception; Encoding unit 23 is used to perform hierarchical encoding of video regions on the real-time video data according to the value score; wherein different value scores correspond to different encoding strategies; Storage unit 24 is used to classify and store the encoded real-time video data based on the value score; wherein different value scores correspond to different storage categories.

[0070] Furthermore, in one possible implementation of this disclosure, the analysis unit 22 is further configured to: A lightweight convolutional neural network model is used to extract spatial features from video frames of the real-time video data, identify specific targets in the video frames in real time, and label the semantic weights of the specific targets. The motion trajectory pattern of the specific target in the real-time video data is captured by a long short-term memory network, the behavioral risk value corresponding to the motion trajectory pattern is calculated, and a behavioral risk prediction value is generated based on a prediction algorithm. By combining the light intensity, weather conditions, and scene sensitivity information in the environmental context data, scene weighting coefficients are generated; The value score is calculated based on the specific target semantic weight, the behavioral risk value, and the scenario weight coefficient.

[0071] Furthermore, in one possible implementation of this disclosure, the encoding unit 23 is further configured to: When the corresponding video region is determined to be a high-value video region based on the value score, lossless encoding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a medium-value video region based on the value score, medium compression coding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a low-value video region based on the value score, the Open Video Coding Standard is used for encoding.

[0072] Furthermore, in one possible implementation of this disclosure, the storage unit 24 is further used for: When a video region is determined to be a high-value video region based on the value score, the video data is stored in the edge node and the high-speed cloud storage medium; wherein, the retention period is set to permanent retention. When the corresponding video area is determined to be a medium-value video area based on the value score, the video data is stored in the local area network additional storage device. When the corresponding video region is determined to be a low-value video region based on the value score, the video data is stored in the edge storage node.

[0073] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 3 As shown, the device further includes: The cleaning unit 25 is used to create a data cleaning task queue in the cache database and perform expired video data cleaning based on the value score and the preset storage period strategy according to the preset period; wherein, the preset storage period is different for video data with different value scores.

[0074] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of the embodiments of this disclosure, and the principle is the same. Therefore, the embodiments of this disclosure are not limited thereto.

[0075] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0076] Figure 4A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0077] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 402 or a computer program loaded from storage unit 408 into RAM (Random Access Memory) 403. RAM 403 may also store various programs and data required for the operation of device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. I / O (Input / Output) interface 405 is also connected to bus 404.

[0078] Multiple components in device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of monitors, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0079] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as video storage optimization methods. For example, in some embodiments, the video storage optimization method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program may be loaded and / or installed on device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, computing unit 401 may be configured to perform the aforementioned video storage optimization method by any other suitable means (e.g., by means of firmware).

[0080] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0081] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0082] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0083] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0084] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0085] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0086] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0087] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0088] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for optimizing video storage, characterized in that, include: Acquire real-time video data and environmental context data, wherein the environmental context data at least includes information on light intensity, weather conditions, and scene sensitivity; Semantic analysis is performed on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data; wherein, the semantic analysis includes spatial feature extraction, temporal behavior analysis, and environmental context perception; Based on the value score, the real-time video data is subjected to hierarchical encoding of video regions; wherein different value scores correspond to different encoding strategies; The encoded real-time video data is classified and stored based on the value score; different value scores correspond to different storage categories.

2. The method according to claim 1, characterized in that, The step of performing semantic analysis on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data includes: A lightweight convolutional neural network model is used to extract spatial features from video frames of the real-time video data, identify specific targets in the video frames in real time, and label the semantic weights of the specific targets. The motion trajectory pattern of the specific target in the real-time video data is captured by a long short-term memory network, the behavioral risk value corresponding to the motion trajectory pattern is calculated, and a behavioral risk prediction value is generated based on a prediction algorithm. By combining the light intensity, weather conditions, and scene sensitivity information in the environmental context data, scene weighting coefficients are generated; The value score is calculated based on the specific target semantic weight, the behavioral risk value, and the scenario weight coefficient.

3. The method according to claim 1, characterized in that, The step of performing hierarchical encoding of video regions on the real-time video data based on the value score includes: When the corresponding video region is determined to be a high-value video region based on the value score, lossless encoding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a medium-value video region based on the value score, medium compression coding is performed using the high-efficiency video coding standard. When the corresponding video region is determined to be a low-value video region based on the value score, the Open Video Coding Standard is used for encoding.

4. The method according to claim 3, characterized in that, The step of classifying and storing the encoded real-time video data based on the value score includes: When a video region is determined to be a high-value video region based on the value score, the video data is stored in the edge node and the high-speed cloud storage medium; wherein, the retention period is set to permanent retention. When the corresponding video area is determined to be a medium-value video area based on the value score, the video data is stored in the local area network additional storage device. When the corresponding video region is determined to be a low-value video region based on the value score, the video data is stored in the edge storage node.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: A data cleanup task queue is created in the cache database, and expired video data is cleaned up based on the value score and the preset storage period strategy according to the preset period; wherein, the preset storage period is different for video data with different value scores.

6. A video storage optimization device, characterized in that, include: The acquisition unit is used to acquire real-time video data and environmental context data, wherein the environmental context data at least covers light intensity, weather conditions and scene sensitivity information; The analysis unit is used to perform semantic analysis on the real-time video data based on a multimodal value perception model to generate a value score for the real-time video data; wherein, the semantic analysis includes spatial feature extraction, temporal behavior analysis, and environmental context perception. The encoding unit is used to perform hierarchical encoding of video regions on the real-time video data according to the value score; wherein different value scores correspond to different encoding strategies; A storage unit is used to classify and store the encoded real-time video data based on the value score; wherein different value scores correspond to different storage categories.

7. The apparatus according to claim 6, characterized in that, The analysis unit is also used for: A lightweight convolutional neural network model is used to extract spatial features from video frames of the real-time video data, identify specific targets in the video frames in real time, and label the semantic weights of the specific targets. The motion trajectory pattern of the specific target in the real-time video data is captured by a long short-term memory network, the behavioral risk value corresponding to the motion trajectory pattern is calculated, and a behavioral risk prediction value is generated based on a prediction algorithm. By combining the light intensity, weather conditions, and scene sensitivity information in the environmental context data, scene weighting coefficients are generated; The value score is calculated based on the specific target semantic weight, the behavioral risk value, and the scenario weight coefficient.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-5.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-5.

Images