A video surveillance storage system and method based on distributed cloud storage

Abstract

This application relates to the field of cloud storage technology. Specifically, it discloses a video surveillance storage system and method based on distributed cloud storage. It collects surveillance video streams through cameras, performs time segmentation on the surveillance video streams based on the video retrieval query status of users to form multiple video data blocks, and then constructs a video distributed cloud storage architecture. It uses the consistent hashing algorithm to determine the storage locations of each video data block in the distributed cloud storage, and monitors the usage frequency of each video data block in real time to dynamically adjust the storage priorities of each video data block. This application can not only effectively avoid the risk of single point of failure, improve the reliability and fault tolerance of storage, but also dynamically adjust the storage priorities of video data according to their usage frequencies, realizing efficient management and access of data.

Description

Technical Field

[0001] This application relates to the field of cloud storage technology, and more specifically, to a video surveillance storage system and method based on distributed cloud storage. Background Art

[0002] With the rapid development of video surveillance technology, the popularity of surveillance cameras has increased year by year, and video surveillance systems are more and more widely used in fields such as public security, traffic management, and commercial operations. However, with the increase in the number of surveillance cameras and the improvement of video resolution, the storage and management of surveillance video data face huge challenges.

[0003] Video surveillance systems usually need to run continuously for 24 hours, generating a huge amount of video data, and traditional centralized storage systems often struggle to meet the storage requirements of large-scale video surveillance data. On the one hand, there is a risk of single point of failure in centralized storage systems. Once the storage center fails, the entire surveillance system will be paralyzed; on the other hand, as the video data continues to increase, the data access efficiency of centralized storage systems will decrease significantly, resulting in increased latency in video retrieval and playback.

[0004] Compared with centralized storage systems, distributed cloud storage systems can not only improve the reliability and fault tolerance of data by dispersing video data storage across multiple nodes, but also enable parallel processing and access of data, further enhancing the efficiency of video retrieval and playback. However, traditional distributed cloud storage systems usually adopt static storage strategies, that is, once video data is assigned to a certain storage node, its storage location remains fixed. This static storage strategy is not flexible enough in the face of frequently changing data access patterns and is difficult to achieve efficient data management and optimization.

[0005] Therefore, an optimized video surveillance storage system and method based on distributed cloud storage are needed to solve the above technical problems. Summary of the Invention

[0006] To solve the above technical problems, this application is proposed. Embodiments of this application provide a video surveillance storage system and method based on distributed cloud storage, which collect surveillance video streams through cameras, perform time chunking on the surveillance video streams based on the video retrieval query status of users to form multiple video data chunks, and then construct a video distributed cloud storage architecture. The consistent hashing algorithm is used to determine the storage locations of each video data chunk in the distributed cloud storage, and the usage frequency of each video data chunk is monitored in real time to dynamically adjust the storage priority of each video data chunk. This can not only effectively avoid the risk of single point of failure, improve the reliability and fault tolerance of storage, but also dynamically adjust the storage priority according to the usage frequency of video data, achieving efficient data management and access.

[0007] According to one aspect of the present application, a video surveillance storage method based on distributed cloud storage is provided, which includes:

[0008] Obtain the surveillance video stream collected by the camera;

[0009] Divide the surveillance video stream into multiple video data blocks;

[0010] Determine the target storage nodes of each video data block in the multiple video data blocks based on the consistent hashing algorithm;

[0011] Based on the target storage nodes of each video data block, store the multiple video data blocks in multiple storage nodes of the distributed cloud storage;

[0012] Monitor the usage frequency of each video data block in the multiple video data blocks, and adjust the storage priority of each video data block based on the usage frequency of each video data block;

[0013] Wherein, monitoring the usage frequency of each video data block in the multiple video data blocks, and adjusting the storage priority of each video data block based on the usage frequency of each video data block includes:

[0014] In response to the usage frequency of the video data block exceeding the first preset threshold, store the corresponding video data block in the storage node with the optimal speed and latency;

[0015] In response to the usage frequency of the video data block being lower than the second preset threshold, store the corresponding video data block in the storage node with the lowest cost in a compressed or archived manner.

[0016] Preferably, the storage node with the optimal speed and latency is an SSD storage node, and the storage node with the lowest cost is an HDD storage node.

[0017] Preferably, dividing the surveillance video stream into multiple video data blocks includes:

[0018] Obtain the user query log, which records all retrieval requests, and the retrieval requests include the time range, camera number, and retrieval result click-through rate;

[0019] Perform semantic embedding encoding on each retrieval request in the user query log to obtain a set of retrieval request semantic embedding encoding vectors;

[0020] Perform demand clustering analysis based on unsupervised learning on the set of retrieval request semantic embedding encoding vectors to obtain video retrieval demand semantic clustering encoding vectors;

[0021] Input the semantic clustering encoded vector of the video retrieval requirement into a decoder-based time-chunk recommender to obtain the recommended values for time chunks;

[0022] Based on the recommended values for time chunks, perform data separation on the monitored video stream to obtain the multiple video data chunks.

[0023] Preferably, perform unsupervised learning-based requirement clustering analysis on the set of semantic embedding encoded vectors of the retrieval request to obtain the semantic clustering encoded vector of the video retrieval requirement, including:

[0024] Input the set of semantic embedding encoded vectors of the retrieval request into an information kernel coarse-grained aggregation network to obtain the coarse-grained semantic aggregation encoded vector of the retrieval request;

[0025] Based on the feature differences between each semantic embedding encoded vector in the set of semantic embedding encoded vectors of the retrieval request and the coarse-grained semantic aggregation encoded vector of the retrieval request, perform dynamic compensation aggregation encoding on the set of semantic embedding encoded vectors of the retrieval request to obtain the fine-grained semantic compensation aggregation encoded vector of the retrieval request;

[0026] Input the fine-grained semantic compensation aggregation encoded vector of the retrieval request and the coarse-grained semantic aggregation encoded vector of the retrieval request into a residual unit to obtain the semantic clustering encoded vector of the video retrieval requirement.

[0027] Preferably, based on the feature differences between each semantic embedding encoded vector in the set of semantic embedding encoded vectors of the retrieval request and the coarse-grained semantic aggregation encoded vector of the retrieval request, perform dynamic compensation aggregation encoding on the set of semantic embedding encoded vectors of the retrieval request to obtain the fine-grained semantic compensation aggregation encoded vector of the retrieval request, including:

[0028] Calculate the kernel aggregation compensation factor of each semantic embedding encoded vector in the set of semantic embedding encoded vectors of the retrieval request relative to the coarse-grained semantic aggregation encoded vector of the retrieval request to obtain a set of semantic kernel aggregation compensation factors of the retrieval request;

[0029] Perform compensation explicit modeling based on a gating function on the set of semantic kernel aggregation compensation factors of the retrieval request to obtain a set of semantic kernel aggregation compensation weight factors of the retrieval request;

[0030] Input the set of semantic kernel aggregation compensation weight factors of the retrieval request, the coarse-grained semantic aggregation encoded vector of the retrieval request, and the set of semantic embedding encoded vectors of the retrieval request into a node fine-grained dynamic compensation aggregation network to obtain the fine-grained semantic compensation aggregation encoded vector of the retrieval request.

[0031] Preferably, calculating a set of retrieval request semantic kernel convergence compensation factors for each retrieval request semantic embedding coding vector in the set of retrieval request semantic embedding coding vectors relative to the retrieval request coarse-grained semantic convergence coding vector, includes:

[0032] Performing point convolution coding based on the Sigmoid activation function on the retrieval request semantic embedding coding vector and the retrieval request coarse-grained semantic convergence coding vector respectively to obtain a normalized retrieval request semantic embedding coding vector and a normalized retrieval request coarse-grained semantic convergence coding vector;

[0033] Calculating a position-wise difference vector between the normalized retrieval request semantic embedding coding vector and the normalized retrieval request coarse-grained semantic convergence coding vector, and taking the absolute value of the position-wise difference vector to obtain a retrieval request semantic kernel convergence difference compensation coding vector;

[0034] Inputting the retrieval request semantic kernel convergence difference compensation coding vector into a compensation feature importance scoring module based on a neural network layer to obtain the retrieval request semantic kernel convergence compensation factor.

[0035] Preferably, inputting the retrieval request semantic kernel convergence difference compensation coding vector into a compensation feature importance scoring module based on a neural network layer to obtain the retrieval request semantic kernel convergence compensation factor, includes:

[0036] Multiplying the retrieval request semantic kernel convergence difference compensation coding vector by a weight parameter matrix, and adding the multiplication result to a bias term to obtain a retrieval request semantic kernel convergence compensation feature modulation vector;

[0037] Multiplying the retrieval request semantic kernel convergence compensation feature modulation vector by a retrieval request semantic compensation feature importance scoring conversion vector to obtain the retrieval request semantic kernel convergence compensation factor.

[0038] Preferably, multiplying the retrieval request semantic kernel convergence difference compensation coding vector by a weight parameter matrix, and adding the multiplication result to a bias term to obtain a retrieval request semantic kernel convergence compensation feature modulation vector, includes:

[0039] Calculating the ratio between the Euclidean norm of the retrieval request semantic embedding coding vector and the Euclidean norm of the retrieval request coarse-grained semantic convergence coding vector, if the ratio is less than 1, then adding 1 to the ratio and calculating the logarithm to the base 2 as the bias term;

[0040] If the ratio is greater than or equal to 1, then taking the ratio as the bias term.

[0041] According to another aspect of the present application, there is provided a video surveillance storage system based on distributed cloud storage, which includes:

[0042] A surveillance video stream acquisition module for acquiring a surveillance video stream collected by a camera;

[0043] A surveillance video stream chunking module for dividing the surveillance video stream into multiple video data chunks;

[0044] A storage node allocation module for determining the target storage node of each video data chunk among the multiple video data chunks based on the consistent hashing algorithm;

[0045] A video data chunk storage module for storing the multiple video data chunks in multiple storage nodes of the distributed cloud storage based on the target storage nodes of each video data chunk;

[0046] A storage priority adjustment module for monitoring the usage frequency of each video data chunk among the multiple video data chunks and adjusting the storage priority of each video data chunk based on the usage frequency of each video data chunk.

[0047] Preferably, the storage priority adjustment module: in response to the usage frequency of the video data chunk exceeding a first preset threshold, stores the corresponding video data chunk in the storage node with the optimal speed and latency; in response to the usage frequency of the video data chunk being lower than a second preset threshold, stores the corresponding video data chunk in the storage node with the lowest cost in a compressed or archived manner.

[0048] The present application has at least the following technical effects:

[0049] Compared with the prior art, the video surveillance storage system and method based on distributed cloud storage provided by the present application collect a surveillance video stream through a camera, perform time chunking on the surveillance video stream based on the video retrieval query status of the user to form multiple video data chunks, and then construct a video distributed cloud storage architecture, use the consistent hashing algorithm to determine the storage location of each video data chunk in the distributed cloud storage, and monitor the usage frequency of each video data chunk in real time to dynamically adjust the storage priority of each video data chunk, which can not only effectively avoid the risk of single point of failure, improve the reliability and fault tolerance of storage, but also dynamically adjust the storage priority according to the usage frequency of video data, realizing efficient management and access of data. Description of the Drawings

[0050] The above and other objects, features, and advantages of the present application will become more apparent by describing the embodiments of the present application in more detail with reference to the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0051] Figure 1 It is a flowchart of a video surveillance storage method based on distributed cloud storage according to an embodiment of the present application.

[0052] Figure 2 It is a flowchart of sub-step S2 of the video surveillance storage method based on distributed cloud storage according to an embodiment of the present application.

[0053] Figure 3 It is a schematic diagram of data flow of sub-step S2 of the video surveillance storage method based on distributed cloud storage according to an embodiment of the present application.

[0054] Figure 4 It is a flowchart of sub-step S23 of the video surveillance storage method based on distributed cloud storage according to an embodiment of the present application.

[0055] Figure 5 It is a flowchart of sub-step S232 of the video surveillance storage method based on distributed cloud storage according to an embodiment of the present application.

[0056] Figure 6 It is a block diagram of a video surveillance storage system based on distributed cloud storage according to an embodiment of the present application. Detailed implementation

[0057] As shown in the present application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "including" and "comprising" only indicate the inclusion of the clearly identified steps and elements, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0058] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0059] In this application, flowcharts are used to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the operations before or below do not necessarily need to be executed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more steps can be removed from these processes.

[0060] Next, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described herein.

[0061] It should be noted that in this application, all actions of obtaining data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining authorization from the corresponding device owner.

[0062] Figure 1 It is a flowchart of a video surveillance storage method based on distributed cloud storage according to an embodiment of this application. As Figure 1 shown, the video surveillance storage method based on distributed cloud storage includes the steps: S1, obtaining a surveillance video stream collected by a camera; S2, dividing the surveillance video stream into multiple video data blocks; S3, determining the target storage nodes of each video data block among the multiple video data blocks based on the consistent hashing algorithm; S4, storing the multiple video data blocks in multiple storage nodes of the distributed cloud storage based on the target storage nodes of each video data block; S5, monitoring the usage frequency of each video data block among the multiple video data blocks, and adjusting the storage priority of each video data block based on the usage frequency of each video data block.

[0063] In the above video surveillance storage method based on distributed cloud storage, in step S1, a surveillance video stream collected by a camera is obtained. Specifically, as a front-end sensing unit, the camera is responsible for capturing the images of the surrounding environment. The captured image information is directly related to the quality of subsequent data processing, storage, and analysis. Therefore, the selection and configuration of the camera are crucial, and the requirements for the camera vary significantly in different application scenarios. For example, in the field of public safety, high-definition cameras can provide clearer images, which helps to identify suspicious behaviors or persons; while in traffic management, panoramic cameras can cover a wider area and effectively monitor road conditions. In addition, with the development of technology, intelligent cameras are gradually becoming popular. Such cameras not only have the traditional image capture function but also have built-in preliminary data analysis capabilities, such as face recognition and behavior analysis. This enables effective data screening at the source and reduces the burden on the back-end system.

[0064] In the above video surveillance storage method based on distributed cloud storage, in step S2, the surveillance video stream is divided into multiple video data blocks. Specifically, since a complete long video stream is not convenient for storage and management. For example, when storing a video, it may require a large amount of continuous storage space, which is not conducive to the flexible allocation of storage resources; when retrieving a video, if one wants to locate the video content in a specific time period, the entire long video needs to be traversed, resulting in low efficiency. Therefore, in this application, by dividing the surveillance video stream into multiple video data blocks, the amount of data stored and retrieved at one time can be effectively reduced, and the flexibility and scalability of data storage can be improved. Among them, Figure 2 is a flowchart of sub-step S2 of the video surveillance storage method based on distributed cloud storage according to an embodiment of this application. Figure 3 is a schematic diagram of data flow of sub-step S2 of the video surveillance storage method based on distributed cloud storage according to an embodiment of this application. As Figure 2 and Figure 3 shown, step S2 includes the steps of: S21, obtaining a user query log, where the query log records all retrieval requests, and the retrieval requests include a time range, a camera number, and a retrieval result click-through rate; S22, performing semantic embedding encoding on each retrieval request in the user query log to obtain a set of retrieval request semantic embedding encoding vectors; S23, performing unsupervised learning-based demand clustering analysis on the set of retrieval request semantic embedding encoding vectors to obtain video retrieval demand semantic clustering encoding vectors; S24, inputting the video retrieval demand semantic clustering encoding vectors into a decoder-based time chunk recommender to obtain recommended values for time chunks; S25, based on the recommended values for time chunks, performing data separation on the surveillance video stream to obtain the multiple video data blocks.

[0065] Specifically, in step S21, a user query log is obtained, and the query log records all retrieval requests, and the retrieval requests include a time range, a camera number, and a retrieval result click-through rate. Specifically, this application takes into account that when dividing the surveillance video stream into chunks, the selection of the time interval will directly affect the subsequent storage and retrieval efficiency. If the time interval is too long, the amount of information contained in each video data block will be too large, and there will still be problems with low storage and retrieval efficiency; if the time interval is too short, it will result in too many video data blocks, increasing the complexity of storage management and user video retrieval. Therefore, in order to improve the convenience of user video retrieval, this application further obtains the user query log, and determines a suitable time interval for video data blocks by analyzing the retrieval request information in the user query log, so that the time period covered by each video data block matches the common time periods in the user retrieval requests, thereby improving the efficiency and accuracy of user video retrieval.

[0066] Specifically, to effectively collect these query logs, it is necessary to consider from the system architecture level to ensure seamless integration into the existing platform. Usually, a complete log management system needs to be established, which consists of multiple components, such as log collectors, log storage servers, and log analysis tools. As the front-end part, the log collector is directly connected to the user interface, responsible for capturing each retrieval request initiated by the user and encapsulating the relevant information into a structured data format. This process requires accurate extraction of parameters such as the time range entered by the user and the selected camera number, and at the same time, it is necessary to track the user's click behavior on the retrieval results to calculate the click-through rate.

[0067] When designing the log collector, considering the stability and reliability in a high-concurrency environment, it is particularly important to adopt an asynchronous processing mechanism. By using a non-blocking data submission method, the system bottleneck problem caused by a large number of log writes can be avoided. In addition, to ensure data consistency and integrity, a unique identifier should be generated every time a retrieval request occurs for subsequent log association and analysis. This not only facilitates tracking the complete query path of a single user but also makes it possible to deeply explore user preferences.

[0068] Regarding the time range dimension, the log collector needs to have flexible parsing capabilities. Whether it is absolute time (such as a specific date and time point) or relative time (such as within the past 24 hours), it should be able to accurately identify and record. For the camera number, since each surveillance camera has a unique identity, it is crucial to clearly record this information in the log. This helps subsequent data analysis based on different camera positions or types, such as counting the surveillance popularity in a certain area or evaluating the working efficiency of a specific device.

[0069] Specifically, in step S22, semantic embedding encoding is performed on each retrieval request in the user query log to obtain a set of retrieval request semantic embedding encoding vectors. Specifically, since the user's retrieval requests are recorded in text form, in order to convert them into a vector form that can be processed by a computer, the present application further uses semantic embedding encoding technology to process each retrieval request in the user query log, so as to map the text information into a high-dimensional semantic space, extract the potential semantic information of each retrieval request, and generate a set of retrieval request semantic embedding encoding vectors, so that the semantic relevance between different retrieval requests can be measured based on the vector distance in the semantic space, and the user's video retrieval access pattern can be understood more accurately. In an embodiment of the present application, the Word2Vec semantic embedding encoding model is used to implement the semantic embedding encoding process of each retrieval request in the user query log.

[0070] Specifically, in step S23, demand clustering analysis based on unsupervised learning is performed on the set of semantic embedding encoding vectors of the retrieval requests to obtain semantic clustering encoding vectors for video retrieval demands. Specifically, in order to extract common features of user video retrievals from a large amount of video retrieval request information, so as to reveal potential video retrieval demands and access patterns of users, this application proposes a demand clustering analysis method based on unsupervised learning. By performing coarse-grained aggregation on the set of semantic embedding encoding vectors of the retrieval requests to construct a global feature summary, and combining a dynamic compensation mechanism, high-fidelity modeling is performed on the detailed semantic features of each retrieval request, so as to generate semantic clustering encoding vectors for video retrieval demands with global semantic structure description ability and local detailed semantic sensitivity. Among them, Figure 4 is a flowchart of sub-step S23 of the video surveillance storage method based on distributed cloud storage according to an embodiment of this application. As Figure 4 shown, step S23 includes the steps of: S231, inputting the set of semantic embedding encoding vectors of the retrieval requests into an information kernel coarse-grained aggregation network to obtain coarse-grained semantic aggregation encoding vectors of the retrieval requests; S232, based on the feature differences between each semantic embedding encoding vector in the set of semantic embedding encoding vectors of the retrieval requests and the coarse-grained semantic aggregation encoding vectors of the retrieval requests, performing dynamic compensation aggregation encoding on the set of semantic embedding encoding vectors of the retrieval requests to obtain fine-grained semantic compensation aggregation encoding vectors of the retrieval requests; S233, inputting the fine-grained semantic compensation aggregation encoding vectors of the retrieval requests and the coarse-grained semantic aggregation encoding vectors of the retrieval requests into a residual unit to obtain the semantic clustering encoding vectors for the video retrieval demands.

[0071] More specifically, step S231 is expressed by the formula:

[0072]

[0073] Among them, represents the set of semantic embedding encoding vectors of the retrieval requests, , , and respectively represent the 1st, 2nd, th, and th vectors in the set of semantic embedding encoding vectors of the retrieval requests, is the number of vectors in the set of semantic embedding encoding vectors of the retrieval requests, and respectively represent taking the maximum value and the minimum value, represents the median of the feature distribution boundary of, represents the normalization exponential function, represents The attention weight represents the coarse-grained semantic aggregation encoding vector of the retrieval request.

[0074] That is, first, the set of semantic embedding encoding vectors of the retrieval request is input into the information core coarse-grained aggregation network. Based on the statistical features of each semantic embedding encoding vector of the retrieval request, its global aggregation weight is calculated to achieve the coarse-grained aggregation encoding of each semantic embedding encoding vector of the retrieval request, thereby extracting the main semantic pattern of the user's retrieval request and obtaining a general expression of the overall semantic characteristics of the set of semantic embedding encoding vectors of the retrieval request, that is, the coarse-grained semantic aggregation encoding vector of the retrieval request.

[0075] Figure 5 is a flowchart of sub-step S232 of the video surveillance storage method based on distributed cloud storage according to an embodiment of the present application. As Figure 5 shown, the step S232 includes the steps of: S2321, calculating the kernel aggregation compensation factor of each semantic embedding encoding vector in the set of semantic embedding encoding vectors of the retrieval request relative to the coarse-grained semantic aggregation encoding vector of the retrieval request to obtain a set of retrieval request semantic kernel aggregation compensation factors; S2322, performing compensation explicit modeling based on a gating function on the set of retrieval request semantic kernel aggregation compensation factors to obtain a set of retrieval request semantic kernel aggregation compensation weight factors; S2323, inputting the set of retrieval request semantic kernel aggregation compensation weight factors, the coarse-grained semantic aggregation encoding vector of the retrieval request, and the set of semantic embedding encoding vectors of the retrieval request into a node fine-grained dynamic compensation aggregation network to obtain the fine-grained semantic compensation aggregation encoding vector of the retrieval request.

[0076] In a specific example of the present application, the step S2321 includes: performing point convolution encoding based on the Sigmoid activation function on the semantic embedding encoding vector of the retrieval request and the coarse-grained semantic aggregation encoding vector of the retrieval request respectively to obtain a normalized retrieval request semantic embedding encoding vector and a normalized retrieval request coarse-grained semantic aggregation encoding vector; calculating the position-wise difference vector between the normalized retrieval request semantic embedding encoding vector and the normalized retrieval request coarse-grained semantic aggregation encoding vector, and taking the absolute value of the position-wise difference vector to obtain a retrieval request semantic kernel aggregation difference compensation encoding vector; inputting the retrieval request semantic kernel aggregation difference compensation encoding vector into a compensation feature importance scoring module based on a neural network layer to obtain the retrieval request semantic kernel aggregation compensation factor. More specifically, inputting the retrieval request semantic kernel aggregation difference compensation encoding vector into a compensation feature importance scoring module based on a neural network layer to obtain the retrieval request semantic kernel aggregation compensation factor includes: multiplying the retrieval request semantic kernel aggregation difference compensation encoding vector by a weight parameter matrix, and adding the multiplication result to a bias term to obtain a retrieval request semantic kernel aggregation compensation feature modulation vector; multiplying the retrieval request semantic kernel aggregation compensation feature modulation vector by a retrieval request semantic compensation feature importance scoring conversion vector to obtain the retrieval request semantic kernel aggregation compensation factor.

[0077] The above step S2321 can be expressed by the formula:

[0078]

[0079] Wherein, represents a compensation factor calculation network, represents an activation function, represents a 1×1 convolution operation, and respectively represent a retrieval request semantic feature weight parameter matrix and a retrieval request coarse-grained semantic aggregation feature weight parameter matrix, represents a normalized retrieval request semantic embedding encoding vector, represents a normalized retrieval request coarse-grained semantic aggregation encoding vector, represents position-wise subtraction, represents a retrieval request semantic kernel aggregation difference compensation encoding vector, represents a retrieval request semantic kernel aggregation difference feature weight parameter matrix, represents a bias term, represents a retrieval request semantic compensation feature importance scoring conversion vector, represents the th retrieval request semantic kernel aggregation compensation factor in the set of retrieval request semantic kernel aggregation compensation factors.

[0080] Specifically, considering that although the above global aggregation method can capture the overall semantic trend of the user's retrieval request, the compressive aggregation process of the global semantic features may dilute or even completely lose the personalized semantic information of some important retrieval requests. Therefore, the present application further introduces the calculation of the kernel aggregation compensation factor, which dynamically generates the kernel aggregation compensation factor of each retrieval request semantic embedding coding vector by measuring the feature deviation between each retrieval request semantic embedding coding vector and the retrieval request coarse-grained semantic aggregation coding vector, and is used to describe the personalized semantic deviation of each retrieval request relative to the overall semantic characteristics, thereby providing strong supplementary support for the subsequent fine-grained compensation modeling of the personalized semantic features of the retrieval requests.

[0081] Particularly, in a preferred example of the present application, multiplying the retrieval request semantic kernel aggregation difference compensation coding vector by a weight parameter matrix and adding the multiplication result to a bias term to obtain a retrieval request semantic kernel aggregation compensation feature modulation vector, including: calculating the ratio between the Euclidean norm of the retrieval request semantic embedding coding vector and the Euclidean norm of the retrieval request coarse-grained semantic aggregation coding vector, if the ratio is less than 1, then calculating the logarithm to the base 2 after adding 1 to the ratio as the bias term; if the ratio is greater than or equal to 1, then using the ratio as the bias term, which is expressed by the formula:

[0082]

[0083] wherein, represents calculating the Euclidean norm of the vector, represents the logarithmic function to the base 2.

[0084] Among them, for the deviation compensation between the retrieval request semantic embedding coding vector and the retrieval request coarse-grained semantic aggregation coding vector, the performance deviation of the kernel aggregation strategy as a scenario strategy can be measured by quantifying the regret metric based on the information kernel compression hypothesis in the kernel aggregation decision process, that is, the game-theoretic counterfactual regret value. Specifically, first, the normalization of the vector norm representation of the counterfactual regret value based on the policy action, that is, the vector norm representation of the retrieval request semantic embedding coding vector and the retrieval request coarse-grained semantic aggregation coding vector, is used as the decision point loss description, and then, for the possible vector distribution action game scenario differences, the compensation rule correction of the retrieval request personalized information is carried out respectively with the regret value information distribution degree and the regret value relative distribution amplitude, so as to consider the retrieval request personalized information as the un-taken action in the decision-making, and perform the bias compensation in the way of assuming its potential benefit based on the information kernel aggregation hypothesis.

[0085] In a specific example of the present application, the step S2322 is expressed by the formula:

[0086]

[0087] Among them, represents the gating threshold, is the natural constant, represents the gating function, represents the th retrieval request semantic kernel convergence compensation weight factor in the set of retrieval request semantic kernel convergence compensation weight factors.

[0088] Specifically, after obtaining the retrieval request semantic kernel convergence compensation factor, the present application further regulates the action magnitude of each retrieval request semantic kernel convergence compensation factor through compensation explicit modeling based on the gating function, and dynamically selects information based on non-linear constraints to screen the action intensity of each retrieval request semantic kernel convergence compensation factor, generating a set of retrieval request semantic kernel convergence compensation weight factors, thereby ensuring that important personalized semantic information is highlighted while irrelevant or redundant information is suppressed.

[0089] In a specific example of the present application, the step S2323 is expressed by the formula:

[0090]

[0091] Among them, represents the retrieval request fine-grained semantic compensation convergence coding vector.

[0092] Specifically, based on the obtained retrieval request semantic kernel convergence compensation weight factor, the difference information (i.e., personalized semantic information) between the retrieval request semantic embedding coding vector and the retrieval request coarse-grained semantic convergence coding vector is compensated and aggregated encoded to generate a retrieval request fine-grained semantic compensation convergence coding vector, thereby realizing the accurate capture and modeling of the personalized semantic features of each user's retrieval request.

[0093] More specifically, the step S233 is expressed by the formula:

[0094]

[0095] Among them, and represent different weight parameters, represents the video retrieval requirement semantic clustering coding vector.

[0096] Specifically, a residual unit is used to perform linear weighted aggregation on the fine-grained semantic compensation aggregation coding vector of the retrieval request and the coarse-grained semantic aggregation coding vector of the retrieval request, so as to combine the global semantic feature generalization expression and personalized semantic information description of the user's video retrieval requirements, generate a semantic clustering coding vector for video retrieval requirements, and thus more accurately describe the potential requirements and access patterns of the user's video retrieval, providing key guidance for subsequent video chunking processing.

[0097] Specifically, in step S24, the semantic clustering coding vector for video retrieval requirements is input into a decoder-based time chunk recommender to obtain a recommended value for time chunking. Specifically, the decoder is based on a multi-layer perceptron architecture. Its input layer is used to receive the semantic clustering coding vector for video retrieval requirements, and through the non-linear transformation of the hidden layer, it deeply understands the user's video retrieval access pattern, and decodes and generates a recommended value for time chunking based on the retrieval access characteristics of the user. The recommended value for time chunking reflects the appropriate size of the video data chunk under the current user's video retrieval access pattern.

[0098] In a preferred example, considering that each retrieval request semantic embedding coding vector in the set of retrieval request semantic embedding coding vectors represents retrieval request semantic embedding coding features, when performing unsupervised learning-based demand clustering analysis on them, since no additional prior information is introduced in unsupervised learning, the semantic clustering coding vector for video retrieval requirements finally obtained by clustering is greatly affected by the noise of local sample points, resulting in local sensitive points or discontinuities at the edges in its high-dimensional feature manifold, affecting the accuracy of its decoding result, that is, affecting the decoding accuracy of the recommended value for time chunking.

[0099] To address the above technical problems, in the technical solution of this application, before inputting the semantic clustering coding vector for video retrieval requirements into a decoder-based time chunk recommender to obtain a recommended value for time chunking, feature manifold optimization is performed on the semantic clustering coding vector for video retrieval requirements to obtain an optimized semantic clustering coding vector for video retrieval requirements, and the process includes the steps of:

[0100] Calculate the phase correlation information between the features at any two positions in the semantic clustering coding vector for video retrieval requirements to obtain a full-granularity mask matrix for the phase of the semantic clustering coding vector for video retrieval requirements, expressed as:

[0101]

[0102] where and represent the features at the th and th positions in the semantic clustering coding vector for video retrieval requirements, represents the first weight hyperparameter, represents the second weight hyperparameter, represents the value at the position in the full - granularity mask matrix of the phase of the semantic clustering encoding for video retrieval requirements;

[0103] Calculate the amplitude correlation information between the features at any two positions in the semantic clustering encoding vector of the video retrieval requirements to obtain the full - granularity mask matrix of the amplitude of the semantic clustering encoding for video retrieval requirements;

[0104]

[0105] wherein, represents the value at the position in the full - granularity mask matrix of the amplitude of the semantic clustering encoding for video retrieval requirements;

[0106] Count the number of valid features in the semantic clustering encoding vector of the video retrieval requirements, and construct a phase search optimization factor and an amplitude search optimization factor based on the number of valid features;

[0107]

[0108]

[0109]

[0110] wherein, , , represent the features at the th, th, th positions in the semantic clustering encoding vector of the video retrieval requirements, represents a preset threshold, for example , which can be adjusted according to the actual situation and is not specifically limited in this embodiment, represents a counting function, represents the number of valid features, represents the phase search optimization factor, represents the amplitude search optimization factor;

[0111] Based on the full - granularity mask matrix of the phase of the semantic clustering encoding for video retrieval requirements and the full - granularity mask matrix of the amplitude of the semantic clustering encoding for video retrieval requirements, and in combination with the phase search optimization factor and the amplitude search optimization factor, perform two - stream enhancement driven by phase search and amplitude search on the semantic clustering encoding vector of the video retrieval requirements to obtain the optimized encoding vector of the phase response of the semantic clustering of the video retrieval request requirements and the optimized encoding vector of the amplitude response of the semantic clustering of the video retrieval request requirements, expressed as:

[0112]

[0113]

[0114] Among them, represents the semantic clustering coding vector of the video retrieval requirement, represents the transpose of the vector, represents matrix multiplication, represents vector addition, represents element-wise multiplication, represents the phase full-granularity mask matrix of the semantic clustering coding of the video retrieval requirement, represents the amplitude full-granularity mask matrix of the semantic clustering coding of the video retrieval requirement, represents the phase response optimized coding vector of the semantic clustering of the video retrieval request requirement, represents the amplitude response optimized coding vector of the semantic clustering of the video retrieval request requirement;

[0115] Fuse the phase response optimized coding vector of the semantic clustering of the video retrieval request requirement and the amplitude response optimized coding vector of the semantic clustering of the video retrieval request requirement to obtain an optimized semantic clustering coding vector of the video retrieval requirement, which is expressed as:

[0116]

[0117] Among them, represents the optimized semantic clustering coding vector of the video retrieval requirement.

[0118] Correspondingly, through the global fine-grained modeling of the phase information and amplitude information at each position of the semantic clustering coding vector of the video retrieval requirement, a phase feature spectrogram and an amplitude feature spectrogram for group aggregation quantization evaluation are constructed, and then an amplitude-phase collaborative enhancement framework based on dual retrieval drive is established based on the phase feature spectrogram and the amplitude feature spectrogram. At the same time, during the collaborative enhancement process, a search optimization factor is constructed based on the effective feature components in the semantic clustering coding vector of the video retrieval requirement to guide the feature optimization to flow in a predetermined direction. Based on the dynamic configuration index response system of the group aggregation phase-amplitude feature topology in the open probability space, combined with the feedback signal fusion strategy, the configuration response redundancy caused by local overload feature parameters is effectively suppressed, and the precise probability fitting of the feature space is finally optimized on the premise of ensuring the feature integrity constraint. In this way, the decoding accuracy of the recommended value of the time block obtained by its input time block recommender based on the decoder is improved.

[0119] Specifically, in step S25, based on the recommended values of the time chunks, the monitoring video stream is segmented to obtain the multiple video data chunks. That is, based on the recommended values of the time chunks, the monitoring video stream is segmented to meet the video retrieval needs of different users, while balancing the integrity of video data and the retrieval speed, optimizing the storage and retrieval efficiency of video data, and ensuring that users can quickly obtain the required content.

[0120] In the above video monitoring storage method based on distributed cloud storage, in step S3, the target storage nodes of each video data chunk among the multiple video data chunks are determined based on the consistent hashing algorithm. Specifically, there are multiple storage nodes in the distributed cloud storage environment. To achieve an even distribution of storage load among the nodes, avoid excessive load on some nodes while some nodes are idle, and at the same time ensure efficient storage and reading of data, this application uses the consistent hashing algorithm to determine the target storage nodes of each video data chunk. The consistent hashing algorithm is a distributed hashing algorithm that can achieve uniform distribution and efficient access of data in a dynamically changing distributed system. Specifically, the consistent hashing algorithm abstracts the hash space as a virtual hash ring and calculates the hash value based on the unique identifier of each storage node in the cloud storage environment (such as IP address, node number, etc.), and maps it to different positions on the hash ring. Then, for the video data chunk to be stored, its hash value is also calculated according to its relevant features (such as the combined information of the data chunk number, timestamp, etc.), and the storage node closest to this hash value is found by clockwise searching on the hash ring as the target storage node. In this way, when the storage nodes increase or decrease, only a small part of the data needs to be migrated to other nodes, rather than causing large-scale data redistribution, thus ensuring that when the nodes change dynamically (such as adding nodes, node failures and exits), the impact on data storage and reading is minimized, and the uniform distribution of data and the dynamic balance of storage load are achieved.

[0121] In the above video monitoring storage method based on distributed cloud storage, in step S4, based on the target storage nodes of each video data chunk, the multiple video data chunks are stored in multiple storage nodes of the distributed cloud storage. That is, each segmented video data chunk is distributed and stored according to its corresponding target storage node to make full use of the storage resources of multiple nodes in the distributed cloud storage system and achieve efficient utilization of storage resources. At the same time, due to the use of the distributed storage method, even if a certain storage node fails, other nodes can still continue to provide services, thus ensuring the continuous and stable operation of the monitoring system. In addition, since multiple storage nodes can respond to data access requests simultaneously to achieve parallel processing and transmission of data, the data access efficiency can be significantly improved, and the delay of video retrieval and playback can be reduced.

[0122] In the above video surveillance storage method based on distributed cloud storage, in step S5, the usage frequency of each video data block among the multiple video data blocks is monitored, and the storage priority of each video data block is adjusted based on the usage frequency of each video data block. Specifically, considering that there are differences in the access frequencies of different video data blocks in actual use. For example, for video data blocks that are frequently accessed, if stored in a storage location with low performance, the reading speed will be slow, affecting the user experience; while for video data blocks that are rarely accessed, if they always occupy high-performance storage resources, it will cause waste of resources. Therefore, based on distributed cloud storage, this application further monitors the usage frequency of each video data block, and adjusts the storage priority of each video data block based on the usage frequency of each video data block, so as to achieve efficient utilization and optimized configuration of storage resources. In a specific example of this application, step S5 includes: in response to the usage frequency of the video data block exceeding a first preset threshold, storing the corresponding video data block in the storage node with the optimal speed and latency; in response to the usage frequency of the video data block being lower than a second preset threshold, storing the corresponding video data block in the storage node with the lowest cost in a compressed or archived manner. More specifically, the storage node with the optimal speed and latency is an SSD storage node, and the storage node with the lowest cost is an HDD storage node.

[0123] In summary, the video surveillance storage method based on distributed cloud storage according to the embodiments of this application is elucidated. It collects surveillance video streams through cameras, performs time segmentation on the surveillance video streams based on the video retrieval query status of users to form multiple video data blocks, then constructs a video distributed cloud storage architecture, uses the consistent hashing algorithm to determine the storage locations of each video data block in the distributed cloud storage, and monitors the usage frequency of each video data block in real time, so as to dynamically adjust the storage priority of each video data block. In this way, not only can the risk of single point of failure be effectively avoided, the reliability and fault tolerance of storage be improved, but also the storage priority of video data can be dynamically adjusted according to the usage frequency of the video data, realizing efficient management and access of data.

[0124] Furthermore, a video surveillance storage system based on distributed cloud storage is also provided.

[0125] Figure 6 It is a block diagram of the video surveillance storage system based on distributed cloud storage according to the embodiments of this application. As Figure 6As shown in the figure, a video surveillance storage system 100 based on distributed cloud storage according to an embodiment of the present application includes: a surveillance video stream acquisition module 110, configured to acquire a surveillance video stream collected by a camera; a surveillance video stream chunking module 120, configured to divide the surveillance video stream into multiple video data chunks; a storage node allocation module 130, configured to determine a target storage node for each of the multiple video data chunks based on a consistent hashing algorithm; a video data chunk storage module 140, configured to store the multiple video data chunks in multiple storage nodes of the distributed cloud storage based on the target storage nodes of each of the video data chunks; a storage priority adjustment module 150, configured to monitor the usage frequency of each of the multiple video data chunks and adjust the storage priority of each of the video data chunks based on the usage frequency of each of the video data chunks. Specifically, the storage priority adjustment module: in response to the usage frequency of the video data chunk exceeding a first preset threshold, store the corresponding video data chunk in the storage node with the optimal speed and latency; in response to the usage frequency of the video data chunk being lower than a second preset threshold, store the corresponding video data chunk in the storage node with the lowest cost in a compressed or archived manner.

[0126] The specific operations of each module in the above video surveillance storage system based on distributed cloud storage have been introduced in detail in the description of the Figures 1 to 5 video surveillance storage method based on distributed cloud storage above, and therefore, the repeated description thereof will be omitted.

[0127] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and easy understanding, rather than limitations. The above details do not limit the present invention to necessarily adopt the above specific details to implement.

[0128] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is merely illustrative. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0129] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0130] In addition, it is obvious that the term "including" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0131] Finally, it should be noted that the above description has been given for purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A video surveillance storage method based on distributed cloud storage, characterized in that, Including: Obtain the surveillance video stream collected by the camera; Divide the surveillance video stream into multiple video data blocks; Determine the target storage nodes of each video data block in the multiple video data blocks based on the consistent hashing algorithm; Store the multiple video data blocks in multiple storage nodes of the distributed cloud storage based on the target storage nodes of each video data block; Monitor the usage frequency of each video data block in the multiple video data blocks, and adjust the storage priority of each video data block based on the usage frequency of each video data block; Dividing the surveillance video stream into multiple video data blocks includes: Obtain the user query log, where the query log records all retrieval requests, and the retrieval requests include the time range, camera number, and retrieval result click-through rate; Perform semantic embedding encoding on each retrieval request in the user query log to obtain a set of retrieval request semantic embedding encoding vectors; Perform unsupervised learning-based demand clustering analysis on the set of retrieval request semantic embedding encoding vectors to obtain video retrieval demand semantic clustering encoding vectors; Input the video retrieval demand semantic clustering encoding vectors into a decoder-based time chunk recommender to obtain the recommended values of the time chunks; Based on the recommended values of the time chunks, perform data separation on the surveillance video stream to obtain the multiple video data blocks; Performing unsupervised learning-based demand clustering analysis on the set of retrieval request semantic embedding encoding vectors to obtain video retrieval demand semantic clustering encoding vectors includes: Input the set of retrieval request semantic embedding encoding vectors into an information kernel coarse-grained aggregation network to obtain retrieval request coarse-grained semantic aggregation encoding vectors; Based on the feature differences between each retrieval request semantic embedding encoding vector in the set of retrieval request semantic embedding encoding vectors and the retrieval request coarse-grained semantic aggregation encoding vectors, perform dynamic compensation aggregation encoding on the set of retrieval request semantic embedding encoding vectors to obtain retrieval request fine-grained semantic compensation aggregation encoding vectors; Input the retrieval request fine-grained semantic compensation aggregation encoding vectors and the retrieval request coarse-grained semantic aggregation encoding vectors into a residual unit to obtain the video retrieval demand semantic clustering encoding vectors.

2. The video surveillance storage method based on distributed cloud storage according to claim 1, wherein Monitoring the usage frequency of each video data block in the multiple video data blocks and adjusting the storage priority of each video data block based on the usage frequency of each video data block includes: In response to the usage frequency of the video data block exceeding the first preset threshold, store the corresponding video data block in a storage node with better speed and lower latency; In response to the usage frequency of the video data block being lower than the second preset threshold, store the corresponding video data block in a storage node with lower cost in a compressed or archived manner.

3. The video surveillance storage method based on distributed cloud storage according to claim 2, wherein The storage node with better speed and lower latency is an SSD storage node, and the storage node with lower cost is an HDD storage node.

4. The video surveillance storage method based on distributed cloud storage according to claim 3, characterized in that, Based on the feature differences between each retrieval request semantic embedding coding vector in the set of retrieval request semantic embedding coding vectors and the retrieval request coarse-grained semantic aggregation coding vector, performing dynamic compensation aggregation coding on the set of retrieval request semantic embedding coding vectors to obtain a retrieval request fine-grained semantic compensation aggregation coding vector, including: Calculating a kernel aggregation compensation factor for each retrieval request semantic embedding coding vector in the set of retrieval request semantic embedding coding vectors with respect to the retrieval request coarse-grained semantic aggregation coding vector to obtain a set of retrieval request semantic kernel aggregation compensation factors; Performing compensation explicit modeling based on a gating function on the set of retrieval request semantic kernel aggregation compensation factors to obtain a set of retrieval request semantic kernel aggregation compensation weight factors; Inputting the set of retrieval request semantic kernel aggregation compensation weight factors, the retrieval request coarse-grained semantic aggregation coding vector, and the set of retrieval request semantic embedding coding vectors into a node fine-grained dynamic compensation aggregation network to obtain the retrieval request fine-grained semantic compensation aggregation coding vector.

5. The video surveillance storage method based on distributed cloud storage according to claim 4, characterized in that, Calculating a kernel aggregation compensation factor for each retrieval request semantic embedding coding vector in the set of retrieval request semantic embedding coding vectors with respect to the retrieval request coarse-grained semantic aggregation coding vector to obtain a set of retrieval request semantic kernel aggregation compensation factors, including: Performing point convolution coding based on the Sigmoid activation function on the retrieval request semantic embedding coding vector and the retrieval request coarse-grained semantic aggregation coding vector respectively to obtain a normalized retrieval request semantic embedding coding vector and a normalized retrieval request coarse-grained semantic aggregation coding vector; Calculating a position-wise difference vector between the normalized retrieval request semantic embedding coding vector and the normalized retrieval request coarse-grained semantic aggregation coding vector, and taking the absolute value of the position-wise difference vector to obtain a retrieval request semantic kernel aggregation difference compensation coding vector; Inputting the retrieval request semantic kernel aggregation difference compensation coding vector into a compensation feature importance scoring module based on a neural network layer to obtain the retrieval request semantic kernel aggregation compensation factor.

6. The video surveillance storage method based on distributed cloud storage according to claim 5, wherein Inputting the retrieval request semantic kernel aggregation difference compensation coding vector into a compensation feature importance scoring module based on a neural network layer to obtain the retrieval request semantic kernel aggregation compensation factor, including: Multiplying the retrieval request semantic kernel aggregation difference compensation coding vector by a weight parameter matrix and adding the multiplication result to a bias term to obtain a retrieval request semantic kernel aggregation compensation feature modulation vector; Multiplying the retrieval request semantic kernel aggregation compensation feature modulation vector by a retrieval request semantic compensation feature importance scoring conversion vector to obtain the retrieval request semantic kernel aggregation compensation factor.

7. The video surveillance storage method based on distributed cloud storage according to claim 6, wherein Multiplying the retrieval request semantic kernel aggregation difference compensation coding vector by a weight parameter matrix and adding the multiplication result to a bias term to obtain a retrieval request semantic kernel aggregation compensation feature modulation vector, including: Calculate the ratio between the Euclidean norm of the semantic embedding encoding vector of the retrieval request and the Euclidean norm of the coarse-grained semantic aggregation encoding vector of the retrieval request. If the ratio is less than 1, calculate the logarithm to the base 2 of the value obtained by adding 1 to the ratio as the bias term; If the ratio is greater than or equal to 1, use the ratio as the bias term.

8. A video surveillance storage system based on distributed cloud storage, for performing the method according to any one of claims 1 to 7, characterized in that, It includes: A monitoring video stream acquisition module for acquiring a monitoring video stream collected by a camera; A monitoring video stream chunking module for dividing the monitoring video stream into multiple video data chunks; A storage node allocation module for determining the target storage node of each video data chunk in the multiple video data chunks based on the consistent hashing algorithm; A video data chunk storage module for storing the multiple video data chunks in multiple storage nodes of a distributed cloud storage based on the target storage nodes of the respective video data chunks; A storage priority adjustment module for monitoring the usage frequency of each video data chunk in the multiple video data chunks and adjusting the storage priority of each video data chunk based on the usage frequency of each video data chunk.

Images