Pushing method, device and equipment of video monitoring product and computer program product

Abstract

The invention provides a video monitoring product pushing method and device, equipment and a computer program product. The method comprises the steps of determining a potential video monitoring demand of a target family based on a family portrait attribute of the target family; and determining at least one target atomic power suitable for the potential video monitoring demand from an atomic power library of a video monitoring service of a computing power network. And configuring a target video monitoring product operated by the computing power network based on a container mirror image created for the at least one target atomic power in advance. And pushing the target video monitoring product to the target family.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, equipment and computer program product for pushing a video surveillance product. Background Art

[0002] The computing power network is "a new type of information infrastructure that allocates and flexibly schedules computing resources, storage resources, and network resources on demand between clouds, networks, and edges based on business needs."

[0003] As people's living standards continue to improve, developing intelligent video surveillance services for families on computing networks has become a future development trend. How to customize the video surveillance products of computing networks for families is the technical problem to be solved by this application. Summary of the invention

[0004] The purpose of this application is to provide a method, device, equipment and computer program product for pushing a video surveillance product, which can customize the video surveillance product of the computing network for the home.

[0005] In order to achieve the above purpose, the embodiment of the present application is implemented as follows:

[0006] In a first aspect, a method for pushing a video surveillance product is provided, comprising:

[0007] Determine the target family's potential video surveillance needs based on their family portrait attributes;

[0008] Determine at least one target atomic capability applicable to the potential video surveillance requirement from an atomic capability library of the video surveillance service of the computing network; the atomic capability refers to a basic independent capability;

[0009] Based on the container image pre-created for the at least one target atomic capability, configuring a target video surveillance product run by the computing power network;

[0010] The target video surveillance product is pushed to the target family.

[0011] In a second aspect, a video surveillance product push device is provided, comprising:

[0012] The demand matching module determines the potential video surveillance needs of the target family based on the family portrait attributes of the target family;

[0013] A capability matching module determines at least one target atomic capability applicable to the potential video surveillance requirement from an atomic capability library of the video surveillance service of the computing power network;

[0014] A product configuration module, configured, based on a container image pre-created for the at least one target atomic capability, to generate a target video surveillance product run by the computing power network;

[0015] The product push module pushes the target video surveillance product to the target family.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; and a memory configured to store computer-executable instructions, wherein the computer-executable instructions, when executed, cause the processor to execute the method described in the first aspect.

[0017] According to a fourth aspect, a computer program product is provided, the computer program product comprising a computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to execute the method described in the first aspect.

[0018] The embodiment of the present application pre-decouples the capabilities of the video surveillance service of the computing power network to build an atomic capability library, and creates corresponding container images for the atomic capabilities in the atomic capability library. When it is necessary to customize video surveillance products for target families, the potential video surveillance needs of the target families are first analyzed based on the family portrait attributes of the target families, and then at least one target atomic capability that matches the potential video surveillance needs is queried from the atomic capability library to configure a target video surveillance product suitable for the target family that is run by the computing power network. In related applications, since the atomic capability container can be reused after being imaged, the computing power network can customize more precise and personalized video surveillance products for different users through the atomic capability library to meet the video surveillance needs of different users. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 This is a first flow chart of the method for pushing a video surveillance product according to an embodiment of the present application.

[0021] Figure 2 This is a second flow chart of the method for pushing a video surveillance product according to an embodiment of the present application.

[0022] Figure 3 This is a third flow chart of the method for pushing a video surveillance product according to an embodiment of the present application.

[0023] Figure 4 This is the first schematic diagram of the LSTM algorithm structure.

[0024] Figure 5 This is the second schematic diagram of the LSTM algorithm structure.

[0025] Figure 6 This is the third schematic diagram of the LSTM algorithm structure.

[0026] Figure 7 This is the fourth schematic diagram of the LSTM algorithm structure.

[0027] Figure 8 A schematic diagram of the family portrait attributes.

[0028] Fig. 9 This is a fourth flow chart of the method for pushing a video surveillance product according to an embodiment of the present application.

[0029] Fig.10 This is a fifth flow chart of the method for pushing a video surveillance product according to an embodiment of the present application.

[0030] Fig.11 This is a schematic diagram of the structure of a push device for a video surveillance product according to an embodiment of the present application.

[0031] Fig.12 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0033] An embodiment of the present application provides a method for pushing a video surveillance product. Figure 1 This is a flowchart of the push method, including:

[0034] S102, determining the potential video surveillance needs of the target family based on the family portrait attributes of the target family.

[0035] In this embodiment, the family portrait attribute of the target family may be obtained by fusing the user portrait attributes of each family member in the target family.

[0036] User portrait attributes may include but are not limited to at least one of basic information, social network, living habits, technology preferences, economic status, and interests and hobbies, which are not specifically limited in this article.

[0037] In one implementation, the weight of each family member of the target family can be determined based on the interaction information between the family members of the target family; then, the user portrait attributes of each family member of the target family are weighted and fused according to the corresponding weight to obtain the family portrait attributes of the target family.

[0038] On the basis of the above, this embodiment can use deep learning technology to realize the prediction of user portrait attributes. That is, a deep learning model is trained based on the user data of sample family members and the user portrait attribute labels (true values) corresponding to the sample family members. During the training process, after the user data of the sample family members are input into the deep learning model, the deep learning model predicts the user portrait attributes of the sample family members as the training result. Then, the loss between the training result and the user portrait attribute label (true value) is calculated, and the model parameters of the deep learning model are adjusted with the purpose of reducing the loss. By iterating the model parameters, the user portrait attributes of the sample family members predicted by the deep learning model can eventually be consistent with the true value, thereby having the ability to predict user portrait attributes. After the deep learning model training is completed, the user data of each family member in the target family can be input into the deep learning model to predict the user portrait attributes of each family member in the target family; after that, the user portrait attributes of each family member in the target family are fused to obtain the family portrait attributes of the target family.

[0039] It should be noted that the deep learning model of this embodiment realizes prediction based on the classification function, and any classification model can be applied to the solution of this embodiment.

[0040] In addition, in order to improve the prediction effect of the deep learning model, in the training stage, the present embodiment can first cluster the user data of the prepared candidate sample family members based on the Gaussian Mixture Model (GMM) clustering algorithm to obtain multiple clusters. Afterwards, it is determined whether each cluster has common user portrait attributes, and if so, the common user portrait attributes are used as user portrait attribute labels of the candidate sample family members corresponding to the cluster, and sample family members are selected from the candidate sample family members with user portrait attribute labels. Through GMM clustering, effective user portrait attributes can be mined to train the deep learning model.

[0041] Among them, the GMM clustering principle is to determine at least two Gaussian distributions corresponding to the user data of the candidate sample family members, and each Gaussian distribution represents a cluster; then, based on the expectation maximization algorithm, the parameters (mean and variance, etc.) of each Gaussian distribution are iterated to determine the cluster to which the user data of the candidate sample family members belongs.

[0042] S104, determining at least one target atomic capability applicable to potential video surveillance needs from the atomic capability library of the video surveillance service of the computing network.

[0043] This embodiment decouples the functions of the video surveillance service of the computing network to obtain multiple basic independent functions, which are atomic capabilities. Then, an atomic capability library is created for the atomic capabilities for users to match. For example, after determining the potential video surveillance needs of the target family, at least one target atomic capability can be matched from the atomic capability library according to the potential video surveillance needs of the target family.

[0044] S106, configuring a target video surveillance product run by the computing power network based on the container image pre-created for at least one target atomic capability.

[0045] This embodiment creates a corresponding container file for each atomic capability in the atomic capability library, and adds the running configuration (such as library files, configuration files, environment variables, network settings, and data volume mapping storage paths, etc.) and running scripts that each atomic capability depends on to the corresponding container file to create a container image for each atomic capability.

[0046] It should be understood that container images are reusable. This embodiment can flexibly select container images with atomic capabilities according to standard design rules of video surveillance products to build different video surveillance products, thereby providing customized video surveillance product services for different users.

[0047] S108, push the target video surveillance product to the target family.

[0048] Specifically, after the target video surveillance product is pushed to the target family, this embodiment can also provide a demonstration of the target video surveillance product for the target family. The target family can provide opinions during the demonstration and adjust the target video surveillance product according to the opinions.

[0049] As an exemplary introduction, a mapping model of a computing power network can be created based on the Systems Modeling Language (SysML), and the mapping model contains the logical relationship between the atomic capabilities and resources of the computing power network; then, based on the mapping model, with the resource requirements of at least one target atomic capability and the resource supply of the computing power network as input parameters, the resource allocation strategy of the computing power network for the target video surveillance product is solved according to the target optimization direction to obtain the target resource allocation strategy; finally, the target resource allocation strategy is provided to the target household, and a demonstration of the target video surveillance product is provided to the target household based on the computing power network according to the target resource allocation strategy.

[0050] Among them, the target optimization direction can be to minimize resource overhead or maximize the target atomic capability performance. For example, first determine a resource allocation strategy according to minimizing resource overhead, and provide a demonstration of the target video surveillance product to the target family according to the resource allocation strategy. If the target family is satisfied, the target video surveillance product is finally implemented by minimizing resource overhead; if the target family is not satisfied, a resource allocation strategy is re-determined by maximizing the target atomic capability performance to provide a demonstration of the target video surveillance product to the target family. In this way, the resources of the algorithm network can be solved as much as possible on the premise of meeting the usage needs of the target family.

[0051] In summary, the method of this embodiment pre-decouples the capabilities of the video surveillance service of the computing power network to build an atomic capability library, and creates corresponding container images for the atomic capabilities in the atomic capability library. When it is necessary to customize video surveillance products for target families, the potential video surveillance needs of the target families are first analyzed based on the family portrait attributes of the target families, and then at least one target atomic capability that matches the potential video surveillance needs is queried from the atomic capability library to configure a target video surveillance product suitable for the target family that is run by the computing power network. In related applications, since the atomic capability container can be reused after being imaged, the computing power network can customize more precise and personalized video surveillance products for different users through the atomic capability library to meet the video surveillance needs of different users.

[0052] The specific application of the method of this embodiment is introduced below.

[0053] The computing network of the method in this embodiment is the artificial intelligence, blockchain, cloud computing, big data, network, edge computing, terminal, security, ABCDNETS computing network. The deep learning model used is the long short-term memory network (LSTM) model. Figure 2 As shown, the process mainly includes the following 9 parts:

[0054] 1) User portrait construction:

[0055] Collect relevant household user information through various channels, use GMM and LSTM to build household user portraits, and match appropriate household labels;

[0056] 2) Potential demand forecast:

[0057] Conduct in-depth analysis of family user portraits and their labels to timely predict the potential video surveillance needs of the family and its members;

[0058] 3) Product demand matching:

[0059] According to the potential business needs of the family, automatically match the corresponding ABCDNETS atomic capabilities (such as computing power, network, AI, etc.) of the atomic capability library of video surveillance products;

[0060] 4) Personalized product design:

[0061] Generate new video surveillance products with the matched ABCDNETS atomic capabilities according to standard product design rules;

[0062] 5) Product solution evaluation:

[0063] Conduct product evaluation on the generated new video surveillance products, including technical feasibility, cost-benefit analysis, etc.;

[0064] 6) Promotion strategy formulation:

[0065] Promptly and accurately promote the evaluated new video surveillance products to relevant home users;

[0066] 7) Product solution implementation:

[0067] Generate a specific resource implementation plan based on the mapping relationship between the atomic capabilities of video surveillance products and related computing network resources and in combination with multiple strategies (such as cost priority and performance priority);

[0068] 8) Product demonstration and trial:

[0069] Provide online demonstrations of the product so that home users can learn about the operation and effects of the product remotely, or provide free trials of the product so that they can experience the advantages of the product firsthand;

[0070] 9) User feedback and product iteration:

[0071] Establish a user feedback mechanism to collect user experience and improvement suggestions for the product, and continuously iterate product functions based on user feedback to improve user satisfaction.

[0072] The process of constructing a family user portrait is as follows: Figure 3 As shown in the figure, we use interest groups, home network devices, web browsing history, social media platforms and other methods to collect multi-source data on home users' video surveillance needs, and perform data backup, cleaning, verification and integration encryption to ensure the quality and security of home user data and build a multi-level, all-round data processing and protection system. Then we use the GMM algorithm to conduct in-depth feature analysis on the integrated home user data and output the final clustering results. According to the user clustering results, we use LSTM to parse the data and match the appropriate family labels to build a family portrait. When new data is collected, it is also processed according to the standard processing flow. Then we use the GMM algorithm to cluster user data, and LSTM parses the clustering results to match the appropriate labels and update them to the previously constructed family portrait.

[0073] The GMM clustering algorithm will first preprocess the data so that the data in each dimension are in the same order of magnitude. This can make the data more stable and reduce the relative deviation between the data. Secondly, a Gaussian model should be established, assuming that the data is composed of several Gaussian distributions. Each Gaussian distribution is a probability distribution function corresponding to one parameter. Therefore, a Gaussian model can be used to describe the distribution of the data. Then, the parameters, mean, and variance of each Gaussian distribution need to be determined, and these parameters are used to calculate the probability distribution of each data point. The maximum likelihood method is then used to determine which Gaussian distribution each data point belongs to. Finally, the expectation maximization (EM) algorithm is used to solve the parameters of the GMM. In the EM algorithm, the model parameters are first initialized, and then the responsiveness of all data points is calculated. Then, the parameters of each Gaussian distribution are updated, and the likelihood function of the model is calculated. Finally, it is determined whether the parameters of the model have converged. If converged, the iteration ends. Otherwise, continue to calculate the responsiveness of all data points and repeat the subsequent steps until the parameters of the model converge.

[0074] The Gaussian Mixture Model (GMM) obtains the probability that each sample point belongs to each class, rather than judging that it completely belongs to one class, which is called soft clustering. When the sample data X is one-dimensional data, the Gaussian distribution follows the following probability density function:

[0075]

[0076] Where u is the data mean (expectation), and σ is the data standard deviation. In GMM, it is assumed that the data is composed of several Gaussian distributions. The probability density function of the Gaussian distribution is:

[0077]

[0078] Where u is the data mean (expectation), Σ is the covariance, and n is the data dimension.

[0079] The GMM algorithm assumes that the data points are generated by one or more Gaussian distributions, and estimates the parameters of the Gaussian distribution of each cluster by the method of maximum likelihood estimation. The Gaussian mixture model can be regarded as a model composed of K single Gaussian models. These K sub-models are the hidden variables of the mixture model. The core idea is to assume that the data can be regarded as generated from multiple Gaussian distributions. Each individual sub-model is a standard Gaussian model with a mean u j and variance Σ j is the parameter to be estimated. In addition, each sub-model has a parameter π j , which can be understood as the weight or probability of generating data. The goal of GMM is to maximize the likelihood function of the data:

[0080]

[0081] Use the expectation maximization (EM) algorithm to solve the parameters of GMM. The process of the EM algorithm is as follows: 1. Initialize the parameters of the model; 2. Calculate the responsiveness of all data points; 3. Update the parameters of each Gaussian distribution; 4. Calculate the likelihood function of the model; 5. Determine whether the parameters of the model have converged. If converged, end the iteration. Otherwise, return to step 2.

[0082] Use the following formula to calculate the responsiveness for each data point:

[0083]

[0084] Use the following formula to update the parameters of each Gaussian distribution:

[0085]

[0086]

[0087]

[0088] Through continuous iteration, the GMM clustering algorithm can be used to cluster the data. After the data is clustered, the LSTM algorithm is used to analyze the data. The structure of the LSTM algorithm is as follows: Figure 4 As shown in the figure, it is a special RNN (Recurrent Neural Network) that can learn long-term dependencies, solve the gradient vanishing and gradient exploding problems in the long sequence training process, and perform better in longer sequences. Unlike RNN, LSTM can remove or add "cell state" information through the "gate" structure, retaining important content and removing unimportant content. The Sigmoid layer outputs a probability value between 0 and 1, describing how much of each part can pass. 0 means "task variables are not allowed to pass", and 1 means "run all variables through". The forget gate, input gate, and output gate control the memory state of the previous information, input information, and output information, thereby ensuring that the network can better learn long-distance dependencies.

[0089] The LSTM forget gate structure is as follows Figure 5 As shown, it is responsible for deciding how much of the cell state at the previous moment to retain in the cell state at the current moment, that is, deciding what information to discard from the cell state. This gate reads h t-1 and x t , and then after passing through the sigmoid layer, output a number f between 0 and 1 t For each cell in state C t-1 Multiply the numbers in point by point. t The value of 0 means completely discarded, and 1 means completely retained.

[0090] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0091] The LSTM input gate structure is as follows Figure 6 As shown in the figure, it is responsible for deciding how much of the current input to keep in the current cell state. It consists of two parts. The first part is the sigmoid layer, which determines what value to update, and the second part is the tanh layer, which updates the information that needs to be updated to the cell state. The tanh layer creates a new cell state value vector Will be added to the state. Then update the old cell state C t-1 C t , the old state and f tMultiply, discard the information that needs to be discarded, and add Complete the update of cell status.

[0092] The LSTM output gate structure is as follows Figure 7 As shown, it is responsible for determining how many outputs the cell state has at the current moment, and a sigmoid layer is used to determine which part of the cell state will be output. The cell state is processed through tanh to obtain a value between -1 and 1, and it is multiplied by the output of the sigmoid gate, and finally only the part that determines the output is output. The structure is shown in the figure below:

[0093] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0094]

[0095] By clustering family user data through GMM and then using LSTM to parse the data, we match the corresponding attribute labels for each family member, and then combine the interaction data between each member to assign individual weights to them, forming an overall family portrait label system, laying the foundation for the precise marketing of personalized video surveillance products. The final family portrait effect is as follows: Figure 8 As shown, it should include at least six dimensions of characteristic information of family members (basic information, social network, living habits, technology preferences, economic status, interests and hobbies), and assign corresponding personal weights to each member based on the interaction information between family members, so as to show a more realistic family status.

[0096] In addition to building an accurate family portrait labeling system, to support the rapid construction of personalized solutions for video surveillance products, it is also necessary to encapsulate the capabilities of video surveillance products according to the ABCDNETS (network, cloud, data, intelligence, security, edge, end, and chain) dimensions, such as Fig. 9As shown. The main implementation steps are: 1. Analyze the functional requirements of the fusion monitoring product, identify the key functional points in each technical field, and determine the technical atomic capabilities that need to be encapsulated; 2. Use containerization technology (such as Docker) to package each atomic capability and its dependencies, and encapsulate the ABCDNETS atomic capability into an independent and reusable service; 3. Deploy the containerized atomic capability to the cloud platform or edge node, and register the service address and metadata of the atomic capability in the service registration center; 4. Deploy the atomic capability to the appropriate operating environment, use container orchestration tools such as Kubernetes to manage the deployment and expansion of the atomic capability, configure load balancing and service discovery mechanisms, and ensure the high availability of the atomic capability; 5. Initiate a call request for the atomic capability through the API gateway. The API gateway routes the request to the corresponding atomic capability service based on the service discovery information to form the overall function of the fusion monitoring product; 6. Use a workflow engine (such as Apache 7. Use monitoring tools (such as Prometheus) to collect performance indicators, monitor the performance of ABCDNETS atomic capabilities, analyze performance bottlenecks, and optimize the service implementation of atomic capabilities; 8. Build an atomic capability library for video surveillance products. The atomic capabilities in the library can be combined in a task-based manner according to customer needs to form new video surveillance products.The most important of these is the second step, which is to encapsulate atomic capabilities in Docker: 1) Determine the containerization target: Identify the specific services or functions that need to be containerized in each technical field (such as AI), and list all the dependencies required for the operation of atomic capabilities, including library files, configuration files, environment variables, etc.; 2) Create Dockerfile: Write a Dockerfile for each atomic capability as a containerization configuration script, and select a suitable base image, such as the official Python or Node.js image; 3) Install dependencies and add code: Use the instructions in the Dockerfile to install the required dependencies, such as scikit-learn, express, and copy the source code of the atomic capability to the Docker container; 4) Configuration and environment settings: Set the necessary environment variables in the Dockerfile, copy the configuration file to the container or pass it through environment variables; 5) Network and storage configuration: Configure the network settings of the container according to the network requirements of the atomic capability. If persistent data is required, use data volumes to map the storage path; 6) Startup command definition: Define the command to run when the container starts in the Dockerfile; 7) Build a container image: Use Docker build command to build a container image according to the Dockerfile; 8) Image testing: Run the container using the built image and test to ensure that everything works properly; 9) Push the image to the image repository: Push the locally built image to Docker Hub or other container image repositories.Through this set of processes, we can build an atomic capability library of video surveillance products. In terms of AI capabilities, video surveillance products can provide AI video analysis capabilities: such as face recognition, license plate recognition, abnormal behavior detection, etc., combined with AI video analysis results, to trigger the corresponding alarm system; in terms of chain capabilities (B), video surveillance products can provide blockchain technology to record monitoring events, ensure that the records cannot be tampered with, and automatically perform monitoring-related tasks, such as automatic alarms, automatic notifications, etc.; in terms of cloud computing (C), video surveillance products can provide remote access services, allowing users to remotely view monitoring videos through the cloud platform, and can also provide cloud storage solutions to support long-term storage and historical playback of videos; in terms of data capabilities (D), video surveillance products support user behavior analysis, by collecting and analyzing user viewing habits, optimize the user experience of video surveillance, and Conduct big data analysis on video content to identify abnormal behavior; in terms of network capabilities (N), video surveillance products are able to transmit data at high speed, support fast and stable transmission of high-definition video streams, and also provide connection options such as wired and wireless; in terms of edge computing (E), video surveillance products support video analysis on cameras or local devices to reduce dependence on the cloud, and also support preliminary processing and storage of videos locally; in terms of terminal capabilities (T), video surveillance products provide multi-terminal compatible services and support viewing video surveillance on multiple devices, including mobile and PC terminals, ensuring compatibility with cameras of different brands and models; in terms of security capabilities (S), video surveillance products support data encryption to ensure that the privacy of transmitted and stored data is protected through encryption technology, and also provide security certification services. Products must pass security certification, such as anti-hacking and anti-cracking.

[0097] With the support of family portraits and product atomic capability libraries, the process of building a personalized video surveillance product solution is as follows: Fig.10As shown. First, potential needs are determined based on family portrait labels and task-based combinations of ABCDNETS atomic capabilities are performed to build video surveillance products of new specifications. Then, relevant resources are scheduled according to the mapping relationship between the atomic capabilities of video surveillance products and cross-ABCDNETS resource pools, and the formulation of demonstration plans and the deployment of relevant resources are quickly completed, which can give users an intuitive product experience and improve user experience and satisfaction. It also supports adjusting product plans based on customer feedback, responding to customer needs in a timely manner, and generating resource deployment plans that meet customer actual needs. The most important part of this process is the rapid design of resource plans. The specific steps are: 1. Resource evaluation and capability analysis: Evaluate the current status of computing network resources, including computing power, storage capacity, network bandwidth, etc. At the same time, analyze the performance requirements and dependencies of atomic capabilities; 2. Design a mapping model: Use system modeling languages ​​(such as SysML) to create a mapping model to define the logical relationship between atomic capabilities and resources. The model should include the type, quantity, performance indicators and cost of resources; 3. Multi-objective optimization: Build an optimization model, take the resource demand of atomic capacity and the supply of computing network resources as variables, and apply multi-objective genetic algorithms to search for the optimal solution; 4. Cloud computing services: According to the load of atomic capacity, use the automatic scaling function of cloud services to adjust resource allocation, such as starting or terminating virtual machine instances; 5. Security integration: Add security scanning steps to the mapping algorithm, use tools such as OWASPZAP to perform security checks on configured resources, and adjust the mapping plan based on the inspection results; 6. Automation script: Write Python scripts to call cloud service APIs, and automatically configure or adjust resources based on the mapping results, such as allocating IP addresses, setting security rules, etc.; 7. Cost-benefit analysis model: Build a cost model, take resource costs, operating costs, expected returns, etc. into consideration, and use financial analysis methods to calculate and compare the net present value (NPV) or internal rate of return (IRR) of different mapping plans; 8. Disaster recovery and backup: Develop data backup and disaster recovery strategies to improve the reliability and fault tolerance of resource plans.When quickly building a resource solution, the capability-resource mapping model built with SysML is indispensable. The model building process is as follows: 1) Create a system overview diagram: Use SysML's System Overview Diagram (SOD) to represent the high-level structure of the system, including the main atomic capabilities and resources, and show the composition and main interactions of the system; 2) Build a block definition diagram: Use SysML's Block Definition Diagram (BDD) to define the blocks in the system (such as AI services, databases, network connections, etc.) and their properties and operations; 3) Describe the relationship between blocks: Use SysML's Internal Block Diagram (IBD) to show the internal structure of the block, as well as the interfaces and dependencies between blocks, to show the connection and interaction between different blocks; 4) Model system behavior: Use SysML's Sequence Diagram (Sequence Diagram) to describe the sequence of events and the interaction between blocks, as well as the State Diagram (State Diagram) to describe the life cycle and state changes of the block, and describe the dynamic behavior of the system, including the activation of atomic capabilities, the allocation and release of resources, etc.; 5) Define requirements and constraints: Use SysML's requirement diagram to express the system's requirements, including performance indicators, safety standards, etc., and use constraint diagrams to express design constraints; 6) Verification and simulation: Use simulation tools to simulate the system behavior according to the SysML model, check whether there are problems or conflicts, and verify the correctness and feasibility of the model; 7) Iteration and optimization: According to the simulation results and feedback from stakeholders, adjust and optimize the model to better meet the needs. Through the above process, the corresponding resource plan can be generated according to the actual needs of customers, and the ABCDNETS resources can be scheduled according to the cost priority and performance priority strategies, making the video surveillance product plan diverse and flexible, and can be personalized according to the label characteristics of different family groups, to maximize the family's actual video surveillance needs, not only to achieve timely response to family video surveillance needs, but also to ensure the efficiency and personalization of the entire service process, to ensure that families can enjoy an efficient and safe one-stop video surveillance experience.

[0098] In summary, the method of this embodiment has the following characteristics:

[0099] 1) Dynamic family portrait label matching system: It integrates multi-dimensional family user information and uses GMM and LSTM to build dynamic user portraits, and matches appropriate attribute labels for each family member and the family as a whole, supporting the generation of corresponding recommendation plans based on family label information;

[0100] 2) Video surveillance product atomic capability library: Use Docker technology to encapsulate the video surveillance product ABCDNETS atomic capabilities and build a cross-domain basic capability library. The capabilities in the library can be combined in a task-based manner according to customer needs to form products of new specifications, and support the upgrade and iteration of atomic capabilities;

[0101] 3) Rapid generation of product solutions: Based on real-time customer feedback, the actual product solution can be quickly generated and the corresponding ABCDNETS resource matching can be completed according to the capability-resource mapping model built by SysML. It supports the generation of recommended selection solutions of different specifications according to customer needs to meet the diverse needs of customers;

[0102] 4) Rapid demonstration of product capabilities: Based on the actual needs of customers, the corresponding product capabilities (such as AI capabilities, alarm capabilities, etc.) can be quickly demonstrated and deployed, allowing customers to have a clear understanding of the required products and improve customer experience and satisfaction.

[0103] Corresponds to Fig.11 The method shown, another embodiment of the present application provides a pushing device for a video surveillance product. Fig.11 1 is a schematic diagram of the structure of the pushing device 1100, comprising:

[0104] The demand matching module 1110 determines the potential video surveillance demand of the target family based on the family portrait attributes of the target family.

[0105] The capability matching module 1120 determines at least one target atomic capability applicable to the potential video surveillance demand from the atomic capability library of the video surveillance service of the computing network; the atomic capability refers to a basic independent capability.

[0106] The product configuration module 1130 configures a target video surveillance product run by the computing power network based on a container image pre-created for the at least one target atomic capability.

[0107] The product push module 1140 pushes the target video surveillance product to the target family.

[0108] The device of this embodiment pre-decouples the capabilities of the video surveillance service of the computing power network to build an atomic capability library, and creates corresponding container images for the atomic capabilities in the atomic capability library. When it is necessary to customize video surveillance products for target families, the potential video surveillance needs of the target families are first analyzed based on the family portrait attributes of the target families, and then at least one target atomic capability that matches the potential video surveillance needs is queried from the atomic capability library to configure a target video surveillance product suitable for the target family that is run by the computing power network. In related applications, since the atomic capability container can be reused after being imaged, the computing power network can customize more precise and personalized video surveillance products for different users through the atomic capability library to meet the video surveillance needs of different users.

[0109] Optionally, before determining the potential video surveillance needs of the target family based on the family portrait attributes of the target family, the demand matching module 1110 further performs: inputting the user data of each family member in the target family into a deep learning model to predict the user portrait attributes of each family member in the target family; wherein the deep learning model is trained based on the user data of sample family members and the user portrait attribute labels corresponding to the sample family members; and fusing the user portrait attributes of each family member in the target family to obtain the family portrait attributes of the target family.

[0110] Optionally, before inputting the user data of each family member in the target family into the deep learning model to predict the user portrait attributes of each family member in the target family, the demand matching module 1110 further performs the following: clustering the user data of the candidate sample family members based on the Gaussian mixture model clustering algorithm to obtain multiple clusters; determining whether each cluster has common user portrait attributes, and if so, using the common user portrait attributes as user portrait attribute labels for the candidate sample family members corresponding to the cluster; and selecting the sample family members from the candidate sample family members with user portrait attribute labels.

[0111] Optionally, the demand matching module 1110 clusters the user data of candidate sample family members based on a Gaussian mixture model clustering algorithm, including: determining at least two Gaussian distributions corresponding to the user data of the candidate sample family members, each Gaussian distribution representing a clustering cluster; iterating the parameters of each Gaussian distribution based on an expectation-maximization algorithm to determine the clustering cluster to which the user data of the candidate sample family members belongs.

[0112] Optionally, the demand matching module 1110 fuses the user portrait attributes of each family member in the target family to obtain the family portrait attributes of the target family, including: determining the weight of each family member of the target family based on the interaction information between the family members of the target family; and weightedly fusing the user portrait attributes of each family member of the target family according to the corresponding weights to obtain the family portrait attributes of the target family.

[0113] Optionally, after configuring the at least one target atomic capability according to the monitoring product design rules into a target video surveillance product running based on the computing power resources of the computing power network, the product push module 1140 further performs: creating a mapping model of the computing power network based on a system modeling language, the mapping model containing the logical relationship between the atomic capabilities and resources of the computing power network; based on the mapping model, taking the resource requirements of the at least one target atomic capability and the resource supply of the computing power network as input parameters, solving the resource allocation strategy of the computing power network for the target video surveillance product according to the target optimization direction to obtain the target resource allocation strategy; the target optimization direction is to minimize resource overhead or maximize the performance of the target atomic capability; providing the target resource allocation strategy to the target household, and providing a demonstration of the target video surveillance product to the target household based on the computing power network according to the target resource configuration strategy.

[0114] Optionally, the device of this embodiment further includes:

[0115] The container configuration module, before configuring the target video surveillance product run by the computing power network based on the container image created in advance for the at least one target atomic capability, performs the following steps: functionally decoupling the video surveillance service of the computing power network to obtain multiple atomic capabilities; creating a corresponding container file for each atomic capability; and adding the running configuration and running script that each atomic capability depends on to the corresponding container file to create a container image for each atomic capability.

[0116] It should be noted that the push method of the video surveillance product of this embodiment can be used as Figure 1 The execution subject of the method shown can thus realize Figure 1 The steps and functions in the method shown are not repeated here.

[0117] Fig.12 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Fig.12At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0118] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.12 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0119] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0120] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming the push device of the above-mentioned video surveillance product at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0121] Based on the family portrait attributes of the target family, determine the potential video surveillance needs of the target family.

[0122] At least one target atomic capability applicable to the potential video surveillance requirement is determined from an atomic capability library of the video surveillance service of the computing power network.

[0123] Based on the container image pre-created for the at least one target atomic capability, a target video surveillance product run by the computing power network is configured.

[0124] The target video surveillance product is pushed to the target family.

[0125] The above application Figure 2The method disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in one or more embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with one or more embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0126] Of course, in addition to software implementation methods, the electronic device of the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0127] The present application also provides a computer program product, which includes a computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the following operations:

[0128] Based on the family portrait attributes of the target family, determine the potential video surveillance needs of the target family.

[0129] At least one target atomic capability applicable to the potential video surveillance requirement is determined from an atomic capability library of the video surveillance service of the computing power network.

[0130] Based on the container image pre-created for the at least one target atomic capability, a target video surveillance product run by the computing power network is configured.

[0131] The target video surveillance product is pushed to the target family.

[0132] In short, the above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of the present application shall be included in the protection scope of one or more embodiments of the present application.

[0133] The systems, devices, and modules described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0134] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0135] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0136] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. A method for pushing a video surveillance product, characterized in that: include: Determine the target family's potential video surveillance needs based on their family portrait attributes; Determine at least one target atomic capability applicable to the potential video surveillance requirement from an atomic capability library of the video surveillance service of the computing network; the atomic capability refers to a basic independent capability; Based on the container image pre-created for the at least one target atomic capability, configuring a target video surveillance product run by the computing power network; The target video surveillance product is pushed to the target family.

2. The method according to claim 1, characterized in that Before determining the potential video surveillance needs of the target family based on the family portrait attributes of the target family, the method includes: Inputting the user data of each family member in the target family into a deep learning model to predict the user portrait attributes of each family member in the target family; wherein the deep learning model is trained based on the user data of sample family members and the user portrait attribute labels corresponding to the sample family members; The user portrait attributes of each family member in the target family are integrated to obtain the family portrait attributes of the target family.

3. The method according to claim 2, characterized in that Before inputting the user data of each family member in the target family into the deep learning model to predict the user portrait attributes of each family member in the target family, the method further includes: Clustering the user data of the candidate sample family members based on the Gaussian mixture model clustering algorithm to obtain multiple clusters; Determine whether each cluster has a common user portrait attribute, and if so, use the common user portrait attribute as the user portrait attribute label of the candidate sample family member corresponding to the cluster; The sample family members are selected from candidate sample family members having user portrait attribute labels.

4. The method according to claim 3, characterized in that: The user data of candidate sample family members are clustered based on the Gaussian mixture model clustering algorithm, including: Determine at least two Gaussian distributions corresponding to the user data of the candidate sample family members, each Gaussian distribution represents a cluster; The parameters of each Gaussian distribution are iterated based on the expectation maximization algorithm to determine the cluster to which the user data of the candidate sample family members belongs.

5. The method according to claim 2, characterized in that: The user portrait attributes of each family member in the target family are integrated to obtain the family portrait attributes of the target family, including: Determining the weight of each family member of the target family based on the interaction information between the family members of the target family; The user portrait attributes of each family member of the target family are weighted and fused according to the corresponding weights to obtain the family portrait attributes of the target family.

6. The method according to any one of claims 1 to 5, characterized in that After configuring the at least one target atomic capability into a target video surveillance product running based on computing power resources of the computing power network according to the monitoring product design rule, the method further includes: Creating a mapping model of the computing power network based on a system modeling language, wherein the mapping model includes a logical relationship between atomic capabilities and resources of the computing power network; Based on the mapping model, taking the resource demand of the at least one target atomic capability and the resource supply of the computing power network as input parameters, solving the resource allocation strategy of the computing power network for the target video surveillance product according to the target optimization direction to obtain the target resource allocation strategy; the target optimization direction is to minimize resource overhead or maximize the performance of the target atomic capability; The target resource configuration strategy is provided to the target household, and a demonstration of the target video surveillance product is provided to the target household based on the computing power network and in accordance with the target resource configuration strategy.

7. The method according to any one of claims 1 to 5, characterized in that Before configuring a target video surveillance product run by the computing network based on a container image pre-created for the at least one target atomic capability, the method further includes: Functionally decoupling the video surveillance service of the computing power network to obtain multiple atomic capabilities; Create a corresponding container file for each atomic capability; Add the running configuration and running script that each atomic capability depends on to the corresponding container file to create a container image for each atomic capability.

8. A push device for a video surveillance product, characterized in that: include: The demand matching module determines the potential video surveillance needs of the target family based on the family portrait attributes of the target family; A capability matching module determines at least one target atomic capability applicable to the potential video surveillance requirement from an atomic capability library of the video surveillance service of the computing power network; the atomic capability refers to a basic independent capability; A product configuration module, configured, based on a container image pre-created for the at least one target atomic capability, to generate a target video surveillance product run by the computing power network; The product push module pushes the target video surveillance product to the target family.

9. An electronic device, comprising: processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to perform a method as claimed in any one of claims 1 to 7.

10. A computer program product, comprising a computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to execute the method according to any one of claims 1 to 7.

Images