WO2023142986A1 - 业务管理方法、平台及服务交付系统、计算机存储介质 - Google Patents

业务管理方法、平台及服务交付系统、计算机存储介质 Download PDF

Info

Publication number
WO2023142986A1
WO2023142986A1 PCT/CN2023/071185 CN2023071185W WO2023142986A1 WO 2023142986 A1 WO2023142986 A1 WO 2023142986A1 CN 2023071185 W CN2023071185 W CN 2023071185W WO 2023142986 A1 WO2023142986 A1 WO 2023142986A1
Authority
WO
WIPO (PCT)
Prior art keywords
service
edge node
transaction
edge
transactions
Prior art date
Application number
PCT/CN2023/071185
Other languages
English (en)
French (fr)
Inventor
王晓红
Original Assignee
京东方科技集团股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 京东方科技集团股份有限公司 filed Critical 京东方科技集团股份有限公司
Priority to US18/282,802 priority Critical patent/US20240168825A1/en
Publication of WO2023142986A1 publication Critical patent/WO2023142986A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5072—Grid computing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/445—Program loading or initiating
    • G06F9/44521—Dynamic linking or loading; Link editing at or after load time, e.g. Java class loading
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/466—Transaction processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/48—Program initiating; Program switching, e.g. by interrupt
    • G06F9/4806—Task transfer initiation or dispatching
    • G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • G06F9/485—Task life-cycle, e.g. stopping, restarting, resuming execution
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45562—Creating, deleting, cloning virtual machine instances
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45575—Starting, stopping, suspending or resuming virtual machine instances
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/50—Indexing scheme relating to G06F9/50
    • G06F2209/502—Proximity
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors

Definitions

  • Embodiments of the present disclosure relate to but are not limited to the technical field of intelligent systems, and in particular relate to a business management method, platform and service delivery system.
  • Edge computing refers to an edge device platform that integrates network, computing, storage, and application core capabilities on the side close to the source of objects or data, and provides the nearest end services. Its applications are launched on the edge side to generate faster network service responses and meet the basic needs of the industry in terms of real-time business, application intelligence, security and privacy protection. Cloud computing can receive or access historical data of edge computing in real time.
  • AI artificial intelligence
  • An embodiment of the present disclosure provides a business management method, including: obtaining monitoring information of edge nodes through a service delivery platform, selecting an available edge node from the obtained monitoring information; creating one or more transactions, and each created transaction It includes one or more artificial intelligence AI services; the created transaction is delivered to the available edge nodes through the service delivery platform.
  • each created transaction further includes at least one of the following: an AI service orchestration script, an image acquisition device corresponding to the transaction, and an available edge node corresponding to the transaction.
  • the method further includes: initializing the available edge nodes.
  • the initializing the available edge nodes includes:
  • the application is a control management program that defines the operation of the edge node
  • the application includes: receiving a delivered transaction, dynamically loading a transaction, parsing a service orchestration transaction, and starting a thread to run a transaction.
  • said creating a transaction includes:
  • the one or more AI services, AI service orchestration scripts, selected image acquisition devices and available edge node information are created as a transaction.
  • the method further includes:
  • the method further includes:
  • the application on the edge node is controlled through the service delivery platform, and the control includes at least one of the following: creating, starting, stopping and updating.
  • the method further includes:
  • the application is a control management program that defines the operation of the edge node
  • a unified shared storage is mounted on the plurality of edge nodes, and the shared storage is configured to store application and transaction configuration data of the plurality of edge nodes.
  • the method when the available edge node fails, the method further includes:
  • the first edge node is an idle edge node, and when there is the first edge node, migrating the applications and transactions of the failed edge node to the first edge node ;
  • the first edge node When the first edge node does not exist, detect whether there is a second edge node, the application version of the second edge node is consistent with the application version of the failed edge node, and the second edge node has a redundant
  • the resource of the transaction of the failed edge node can be received, and when the second edge node exists, the association relationship between the application and the transaction of the failed edge node is deleted, and the transaction of the failed edge node is migrate to said second edge node;
  • alarm information is generated.
  • the method further includes:
  • the one or more available edge nodes are obtained through any one or more of the following service channels: message queue telemetry transport protocol, remote dictionary service, distributed publish-subscribe message system, and network socket AI service processing results.
  • one edge node runs multiple transactions in parallel, and each transaction is started as a thread.
  • An embodiment of the present disclosure also provides a service management platform, including a memory; and a processor coupled to the memory, the processor is configured to execute the operation described in any one of the above based on instructions stored in the memory. Steps of the business management method described above.
  • An embodiment of the present disclosure also provides a service delivery system, including the above-mentioned service management platform, and also includes a service delivery platform and one or more edge nodes, the service management platform, service delivery platform and edge nodes communicate with each other through the network connect;
  • the service delivery platform is configured to monitor and control the one or more edge nodes, and send the transaction created by the service management platform to the one or more edge nodes;
  • the edge node is configured to receive the transaction issued by the service delivery platform, process the transaction, and return the processing result to the service management platform
  • the service delivery system further includes one or more image acquisition devices, and the edge node and the image acquisition devices are connected to each other through a network;
  • the processing of the transaction by the edge node includes: acquiring corresponding image or video data collected by one or more image collection devices according to the received transaction, and processing the image or video data.
  • the edge node runs multiple transactions in parallel, and the transactions include customer face registration transactions, customer identification transactions, store visit frequency analysis transactions, and stranger reminder transactions, wherein:
  • Customer face registration services include face detection service, face zoom service, face key point detection service, face alignment service, face feature extraction service and face feature storage service;
  • Customer identification services include video decoding service, image scaling service, target detection service, target tracking service, face detection service, face scaling service, face quality evaluation service, face alignment service, face attribute judgment service, face feature extraction service, face retrieval service and stranger registration service;
  • the visit frequency analysis service includes record upload service and record statistics service
  • the stranger reminder service includes the statistics service of strangers' visits to the store and the reminder service.
  • An embodiment of the present disclosure also provides a computer storage medium, on which a computer program is stored, and when the program is executed by a processor, the service management method described in any one of the preceding items is implemented.
  • FIG. 1 is a schematic flowchart of a business management method in an exemplary embodiment of the present disclosure
  • FIG. 2 is a schematic flow diagram of edge node initialization according to an exemplary embodiment of the present disclosure
  • FIG. 3 is a schematic flow diagram of transaction issuing/updating in an exemplary embodiment of the present disclosure
  • FIG. 4 is a schematic diagram of multiple AI service linkages in a vehicle violation detection transaction according to an exemplary embodiment of the present disclosure
  • 5a and 5b are schematic structural diagrams of two service delivery systems according to exemplary embodiments of the present disclosure.
  • FIG. 6 is a schematic diagram of an application scenario of a store visit frequency detection transaction according to an exemplary embodiment of the present disclosure
  • FIG. 7 is a schematic diagram of a customer identification process in a store visit frequency detection transaction according to an exemplary embodiment of the present disclosure
  • FIG. 8 is a schematic diagram of a customer visit frequency detection process in a store visit frequency detection transaction according to an exemplary embodiment of the present disclosure
  • Fig. 9 is a schematic structural diagram of a service management platform according to an exemplary embodiment of the present disclosure.
  • the embodiment of the present disclosure provides a service management method, including the following steps:
  • Step 101 Obtain monitoring information of edge nodes through the service delivery platform, and select available edge nodes from the obtained monitoring information;
  • the business management method of the embodiment of the present disclosure is applied on the business management platform, the business management platform is the service initiator, the edge node is the operation unit of the transaction, and the service delivery platform controls the edge node through the cloud network.
  • the edge node is pre-installed with edge components (Edge part) at the time of delivery, and becomes a K8S node (Node).
  • the core application (APP) of the edge node can be delivered through Kubernetes.
  • Kubernetes (K8S for short) is a brand-new distributed architecture solution based on container technology and an open source container cluster management system.
  • a Kubernetes cluster generally includes a control (Master) node and multiple Node nodes.
  • the Master node is the cluster control node of K8S.
  • Each K8S cluster needs a Master node to be responsible for the management and control of the entire cluster. Basically, all control commands of K8S are sent to it, and it is responsible for the actual execution process.
  • the Node node is the workload node in the K8S cluster. Each Node node will be assigned some workload by the Master node. When a Node node goes down, the workload on it will be automatically transferred to other Node nodes by the Master node.
  • the method further includes: initializing available edge nodes.
  • initializing available edge nodes includes:
  • APP an application
  • the application is a control management program that defines the operation of the edge node
  • the application container image is constructed through the service delivery platform, and the built application container image is stored in the container warehouse, and the available edge node is controlled by the service delivery platform to download the application container image and build and start the application container.
  • the business management platform selects the application version according to the version of the application currently required; the business management platform obtains the basic container image list from the service delivery platform, and selects the basic container image from the basic container image list; the business management platform Obtain edge node monitoring information from the service delivery platform, and select available edge nodes from the obtained edge node monitoring information; the business management platform builds an application container image through the service delivery platform based on the selected application and basic container image and The application container image is stored in the container warehouse, and the available edge nodes are controlled by the service delivery platform to download the application container image and build and start the application container.
  • the basic container image may be a Linux operating system such as Redhat (Redhat), Ubuntu (Ubuntu), etc., and the basic container image may provide a basic environment for application operation.
  • Redhat Redhat
  • Ubuntu Userbuntu
  • the application may include: receiving delivered transactions, dynamically loading transactions, parsing service orchestration, starting threads to run transactions, and the like.
  • the application may also include: video decoding transactions, data stack management transactions, and the like.
  • the application may include a video decoding transaction and a data stack management transaction, wherein the video decoding transaction is responsible for decoding the received video, and the data stack management transaction is responsible for storing the decoded image data.
  • the edge node directly receives the image data through the network, the application may not include the video decoding transaction and the data stack management transaction.
  • the available edge node starts the above application, that is, it has the ability to receive delivered transactions, dynamically load transactions, analyze service orchestration, start thread running transactions, and so on.
  • Step 102 create one or more transactions, each created transaction includes one or more AI services;
  • each created transaction further includes at least one of the following: an AI service orchestration script, an image acquisition device corresponding to the transaction, and an available edge node corresponding to the transaction.
  • creating a transaction includes:
  • AI service code bases Create one or more AI service code bases, AI service orchestration scripts, selected image acquisition devices, and available edge node information as a transaction.
  • the business management platform selects one or more AI services, and generates an AI service orchestration script based on the selected one or more AI services; the business management platform selects one or more Image acquisition equipment; the business management platform obtains edge node information from the service delivery platform, and selects available edge nodes from the obtained edge node information; the business management platform packages one or more AI service code libraries and AI service orchestration scripts, and distributes them to the available edge nodes; the business management platform obtains the transaction delivery or update status of the edge nodes through the service delivery platform.
  • the edge node receives the delivered transaction, it downloads the transaction package, dynamically loads the AI library, parses the service orchestration script, creates the AI service call sequence, and then starts the transaction thread.
  • an edge node runs multiple transactions in parallel, and each transaction is started as a thread.
  • AI technologies such as speech recognition and image recognition
  • a single AI algorithm model is often only for a single problem.
  • a transaction often requires the linkage of multiple AI services, such as As shown in Figure 4, taking illegal parking monitoring as an example, if a vehicle picture is input into the model recognition model, the model can be identified, and if the model needs to be identified, different AI services need to be input.
  • application scenarios often require data sources such as cameras to solve scenarios such as illegal parking monitoring.
  • the AI service orchestration process can be drawn through a web page or other visual means, and a corresponding AI service orchestration script can be generated according to the flow chart to realize linkage of multiple AI services.
  • Step 103 send the created transaction to one or more available edge nodes corresponding to the transaction through the service delivery platform.
  • the method also includes:
  • the edge node After the edge node receives the delivered transaction, it first downloads the transaction package, then dynamically loads the AI library, parses the service orchestration script and creates the AI service call sequence, and then starts the transaction thread. After the transaction is successfully started, the edge node is updated. node status.
  • the method also includes:
  • the application on the edge node is controlled through the service delivery platform, and the control includes at least one of the following: creating, starting, stopping and updating.
  • the service delivery platform can be designed and developed based on K8S KubeEdge, edge nodes are installed with Edge part to manage and run APP containers, and the service delivery platform manages the life cycle of APP containers (creation, start, stop, update, etc.) through Edge part.
  • the node control depends on K8S’ ability to control the container, so it needs to be placed on the service delivery platform. If it is placed on the business management platform, that is, the business management platform directly controls the edge nodes, it needs to be customized Own control (creation, update, start, stop) logic and protocol, which will increase the development workload.
  • the method also includes:
  • message queue telemetry transport protocol Message Queuing Telemetry Transport, MQTT
  • remote dictionary service Remote Dictionary Server, Redis
  • distributed publish and subscribe message system such as Kafka
  • WebSocket a full-duplex communication protocol based on Transmission Control Protocol (TCP)
  • TCP Transmission Control Protocol
  • the business management method provided by the embodiment of the present disclosure implements a cloud-native edge node management method through the business management platform scheduling service delivery platform for transaction delivery, container image construction, edge node monitoring, etc., and the transaction delivery method is efficient and convenient ,
  • the transaction management mechanism is efficient and flexible, and through AI service orchestration, the flexible construction, convenient and controllable transaction is guaranteed.
  • An embodiment of the present disclosure also provides a service delivery system, including the aforementioned service management platform, a service delivery platform, and one or more edge nodes.
  • the business management platform, the service delivery platform and the edge nodes are connected to each other through a network, and the edge node and the image acquisition device are connected to each other through a network.
  • the service delivery platform is configured to monitor and control one or more edge nodes, and deliver the transactions created by the business management platform to one or more edge nodes;
  • the edge node is configured to receive the transaction issued by the service delivery platform, process the transaction, and return the processing result to the business management platform.
  • the service delivery system further includes one or more image acquisition devices, and the edge nodes and the image acquisition devices are connected to each other through a network;
  • the edge node processes the transaction, including: according to the received transaction, obtains image or video data collected by corresponding one or more image collection devices, and processes the image or video data.
  • the service delivery system provided by the embodiment of the present disclosure at least includes the following components: a service management platform, a service delivery platform and an edge node.
  • the business management platform is the service initiator, and its functional modules include but not limited to edge node management module, camera management module, transaction management module, APP management module, AI service management module, AI service display module, etc.
  • Edge node management module responsible for configuring all edge node information of the business management platform, and interacting with the edge monitoring module of the service interaction platform to obtain the resource usage of the current edge node.
  • Camera management module responsible for recording the configuration information of all camera video streams (for example, the configuration information includes information such as stream address, camera model, camera position, camera manufacturer, etc.), and the association relationship between each camera and the transaction (that is, the data of the transaction which camera or cameras the source comes from).
  • Transaction management module Define an AI service capability as a transaction. To create a transaction, you need to configure AI services, generate AI service orchestration scripts, select cameras, select available edge nodes, etc. In Figure 5, the frequency of arrivals or departure detection is The example defines a transaction that needs to interact with other modules to obtain the necessary configuration information.
  • APP management module refers to the control management program that defines the operation of edge nodes.
  • the main functions of APP are video stream decoding, AI service management and scheduling, etc.
  • APP management includes starting APP, stopping APP, updating APP, etc.
  • AI service management module This module is responsible for packaging AI services into modules that can be scheduled by the APP, and defines the linkage arrangement relationship between multiple AI services. In the embodiments of the present disclosure, how to define the linkage orchestration between AI services needs to be defined according to actual usage scenarios, which is not limited in the present disclosure.
  • AI service display module responsible for displaying the processing results of AI services from edge nodes.
  • the service delivery platform can be designed and developed based on K8S KubeEdge.
  • the Edge part is installed on the edge to manage the edge container running the APP.
  • the service delivery platform manages the life cycle of the APP container (creation, start, stop, delete, etc.) through the Edge part, and through
  • the Open API of the service delivery platform is dispatched by the business management platform to the service delivery platform to build APP container images, update transactions, control edge nodes, and return edge node monitoring information to the business management platform.
  • KubeEdge is Kubernetes' native edge computing platform.
  • the KubeEdge architecture includes two parts, namely the cloud and the edge side.
  • the cloud is responsible for delivering applications and configurations, and the edge side is responsible for running edge applications and managing access devices.
  • the edge node is the operation unit of the transaction, and the edge node is pre-installed with the Edge part at the time of delivery, and becomes a K8S Node node.
  • the core application APP of the edge node is delivered through K8S.
  • APP applications mainly include video decoding transactions, data stack management transactions, receiving and sending transactions, dynamic loading transactions, parsing service orchestration transactions, starting thread running transactions, etc.
  • An edge node can run multiple transactions in parallel, each transaction can be started as a thread, and different transactions can be executed in parallel.
  • an edge node runs two transactions: transaction 1 and transaction 2.
  • transaction 1 can be used to detect human faces;
  • transaction 2 can be used to detect human bodies , since transaction 1 and transaction 2 may have different requirements for the shooting accuracy and shooting range of the camera, separate detection is beneficial to control the shooting accuracy and shooting range of the corresponding camera separately.
  • the same detection target can be The detection results of transaction 1 and transaction 2 are packaged and sent back to the business management platform.
  • the results of the transaction are reported to the business management platform by the edge nodes through any one or more of the following service channels: message queue telemetry transmission protocol, remote dictionary service, distributed publish and subscribe message system, network socket, etc.
  • the results output by the edge AI transaction may include information such as display, alarm, and notification.
  • the AI service processing result of the edge node finds that there is no relevant personnel in the current position, it will send information such as (a certain position: leaving the job) to the AI service display module of the business management platform, and the AI service display module An alarm box can pop up "a certain person leaves the post" and send an alarm to the relevant management personnel.
  • the screen processed by the AI service can be pulled through the WebSocket protocol and other methods to display the information and location information of off-duty personnel.
  • the AI service processing result returns information such as the customer ID and arrival time to the AI service display module, and the AI service display module receives the customer's information , first record the customer's current store visit information to the database, and then count the customer's store visit times in a specified time period (such as 1 year), and you can pop up the 1-year store visit times > preset store visits on the web page (Web) side
  • the prompt information of the number of times threshold (such as 5 times), so that the clerk can focus on the customer.
  • the threshold value of the preset number of visits to the store can be adjusted according to needs. For example, when you need to pay attention to unfamiliar customers, you can pop up the information of the first-time customers.
  • the size can be configured through the AI service display module.
  • edge node 1 When the service delivery platform monitors that a certain edge node (such as edge node 1) fails (such as downtime), it will report the fault information to the edge node management module of the business management platform for processing.
  • the node runs the application (APP) container and the issued transaction at the same time, so it can be divided into the following two situations to achieve high availability:
  • APP application
  • edge node 2 When there is an idle edge node (such as edge node 2), since the application is packaged into an application container (docker) image, the application can be migrated to other available idle edge nodes (such as edge Node 2). However, the transaction data cannot be directly migrated from the failed edge node 1 to the idle edge node 2, therefore, in some exemplary embodiments, in the transaction management module of the service management platform, the application and Transaction association, and a unified shared storage is mounted on multiple edge nodes, and the shared storage is used to store application transaction configuration data of multiple edge nodes. Therefore, when migrating the application and transaction data of edge node 1 to edge node 2, it is only necessary to read the application transaction configuration data of edge node 1 when edge node 2 starts the application.
  • the relationship between the application and the transaction of the edge node can be expressed as APP-ID1 (transaction 1, transaction 2), where ID1 of APP-ID1 can be the name of the docker image packaged by the application, and the ID can be used to associate This application and transaction.
  • the edge node management module deletes the relationship between the application and the transaction of the edge node that has failed in the transaction management module.
  • Delete the relationship between APP-ID1 and transactions 1 and 2 on the shared storage so as to avoid multiple executions of one transaction when edge node 1 is repaired, and send the corresponding transactions 1 and 2 to other available edge nodes of the same version through the transaction management module (such as edge node 2), update the association relationship between edge node applications and transactions in shared storage, such as configuring APP-ID2 (transaction 4, transaction 1, transaction 2), and update the association between edge node applications and transactions in the transaction management module Relationships, such as record APP-ID2(transaction4, transaction1, transaction2).
  • the edge node management module triggers an alarm mechanism, and sends text messages, WeChat, emails, etc. to trigger manual intervention processing.
  • the edge monitoring module of the service delivery platform When a certain camera fails and the application of the edge node performs video encoding and decoding, it finds that the program cannot run normally and sends an alarm message. After the edge monitoring module of the service delivery platform receives the information, the edge monitoring module will feed back the received information to the The edge node management module of the business management platform triggers the alarm mechanism, and sends text messages, WeChat, emails, etc. to trigger manual intervention processing.
  • the edge monitoring module is placed on the service delivery platform because K8S monitoring systems such as Prometheus (an open source system monitoring and alarm system) can be used directly.
  • K8S monitoring systems such as Prometheus (an open source system monitoring and alarm system) can be used directly.
  • the business management platform focuses more on business capability management, while the service The delivery platform focuses on delivery operation and maintenance management.
  • the business management method of the present disclosure can complete transaction hot update.
  • update transaction 2 As an example, firstly, through the transaction management module of the business management platform, call the transaction update module of the service delivery platform, and notify the application APP-ID1 of edge node 1 to delete transaction 2.
  • APP-ID1 passes the main thread Stop exiting the transaction 2 thread, delete the edge transaction configuration data APP-ID1 (transaction 1, transaction 2) of the node on the centralized shared storage, and change it to APP-ID1 (transaction 1), and the edge node reports to the transaction update module of the service delivery platform
  • the deletion task has been completed, and then the transaction management module of the business management platform regenerates the transaction delivery.
  • the delivered edge node does not need to be selected, and it is still edge node 1. Start the transaction delivery and transaction start process, and update the transaction management module.
  • the embodiments of the present disclosure design a customer visit frequency statistics system based on edge computing, which can provide a safe, efficient, and intelligent customer visit statistics method.
  • the system includes cameras, networks, and edge devices (i.e., edge nodes) in hardware, and video stream processing modules, target detection modules, target tracking modules, face correction modules, face recognition modules, etc. in software, as well as data storage,
  • edge devices i.e., edge nodes
  • the deployment architecture is shown in Figure 6 for functional modules such as analysis calculation, display, and cloud synchronization.
  • the edge device by deploying the edge device in the store or close to the store, and the edge device is close to the application scene, low-latency video transmission can be realized, and high-real-time and visible customer arrival records or reminders can be realized.
  • the system consists of two parts: software and hardware:
  • Hardware includes: image acquisition devices (such as cameras), image processing devices (such as edge nodes), network devices that can connect image acquisition devices and image processing devices, service delivery platforms, and business management platforms.
  • image acquisition devices such as cameras
  • image processing devices such as edge nodes
  • network devices that can connect image acquisition devices and image processing devices
  • service delivery platforms such as business management platforms.
  • Software includes: The software of this system mainly runs on edge nodes, including video decoding module, target detection module, target tracking module, face correction module, face recognition module, etc.
  • the “module” mentioned in this disclosure may also be called “service”. Based on the processing results of each module, the customer's visits to the store are stored, and the frequency of visits to the store is analyzed and displayed on a daily, weekly, monthly, and annual basis.
  • the target detection module is configured to find out the target from a scene (picture), including two processes of detection (where) and recognition (what).
  • the object tracking module is configured to establish the positional relationship of the object to be tracked in the continuous video sequence, and obtain the complete motion track of the object.
  • the target tracking module calculates the exact position of the target in the next frame of the image according to the target coordinate position of the first frame of the given image.
  • the target may show some changes in the image, such as changes in posture or shape, changes in scale, background occlusion, or changes in light brightness.
  • Object tracking technology is one of the hotspots in the field of computer vision research and has been widely used. The tracking and focusing of the camera, the automatic target tracking of the drone, etc. all need to use the target tracking technology.
  • there are specific object tracking such as human body tracking, vehicle tracking in traffic monitoring systems, face tracking and gesture tracking in intelligent interactive systems, etc.
  • the face correction module is configured to, when the angle of the detected face is not very straight, align it through methods such as face key point detection and key point rotation transformation.
  • the face recognition module includes a face recognition submodule and a face verification submodule, and the face recognition submodule is configured to classify a face as a specific identification (identification); the face verification submodule is configured to determine Whether a pair of pictures belong to the same person (Verification).
  • multiple transactions can be run in parallel on one edge node.
  • the multiple transactions include customer face registration transactions, customer identification transactions, store visit frequency analysis transactions, and stranger reminder transactions, wherein:
  • Customer face registration services include face detection service, face zoom service, face key point detection service, face alignment service, face feature extraction service and face feature storage service;
  • Customer identification services include video decoding service, image scaling service, target detection service, target tracking service, face detection service, face scaling service, face quality evaluation service, face alignment service, face attribute judgment service, face feature extraction service, face retrieval service and stranger registration service;
  • the visit frequency analysis service includes record upload service and record statistics service
  • the stranger reminder service includes the statistics service of strangers' visits to the store and the reminder service.
  • system includes but is not limited to the following functions:
  • the system needs to first complete the face registration based on the customer's picture.
  • the management personnel can upload the picture through the interface provided by the edge node, such as a web page (Web) service. After the picture is uploaded, the customer registration function is triggered. First pass the picture to the target detection module (this module contains a face detection algorithm) to identify the customer's face information.
  • the target detection module this module contains a face detection algorithm
  • the face may be tilted and other situations affect the accuracy of customer recognition.
  • the alignment process includes first scaling the face size and calling the face key
  • the point detection algorithm identifies the key point information of the face, which is passed into the face alignment algorithm for face alignment to achieve head tilt correction.
  • call the face recognition module to extract face features
  • the face storage module to store the extracted features in the edge node.
  • the customer image can be deleted after extracting face features.
  • the embodiments of the present disclosure store the user data in the edge device, and only store the facial features of the customer, so as to ensure the security and privacy of the customer data.
  • three cameras are taken as an example to collect data, so as to prevent one camera from affecting the user's face data extraction due to reasons such as occlusion.
  • the input video stream is first decoded by the video decoding module, the decoded image is scaled and the target detection module is called to detect the pedestrian information in the image, and the target tracking module is used to judge whether the pedestrian appears for the first time.
  • the target tracking module judges that the pedestrian appears for the first time, face detection is performed on the pedestrian, and the image is zoomed after the face is detected, and input to the face quality evaluation module to judge whether the face is blurred or blocked. If it is qualified, return to the next round of video stream processing; when the quality of the face is qualified, perform face alignment to correct the tilted head, etc., input the corrected result into the face attribute module, obtain auxiliary information such as age and gender, and then call the human
  • the face feature extraction module extracts face features, uses the extracted features to search for matching customers in the face feature database, and determines the final customer after comprehensive matching with face attribute features, binds the pedestrian with the customer, and records Customer arrival time and customer trajectory information.
  • the target tracking module judges that the pedestrian is not the first time, it judges whether the pedestrian is bound to the customer.
  • the pedestrian is bound to the customer, record the customer's arrival time and the customer's trajectory information, otherwise as a stranger, record the customer's arrival at the store time, and then return for the next round of video stream processing.
  • the embodiment of the present disclosure can more accurately capture the real-time information of the customer's arrival and record the customer's arrival and departure through technologies such as multiple cameras, target tracking, and accurate face recognition, and provide the reference for the operator.
  • Customer arrival records and in-store trajectory information can be stored on the edge nodes, which can ensure the security of sensitive data, but problems such as equipment damage may cause data loss. Therefore, customers can be transferred while ensuring data security (such as encrypted transmission).
  • Store arrival records and in-store track information are uploaded to the cloud.
  • this disclosure also records the customer's in-store track information, which can analyze the customer's track in the store as needed, and provide some operators with reference to customer portraits.
  • This disclosure designs a stranger face database for strangers.
  • the stranger When the target pedestrian is not a registered customer, the stranger’s face database is searched.
  • the target pedestrian When the target pedestrian is a non-registered stranger, first register and then record the time to the store and departure time. The completion of the record triggers the statistics of strangers' arrivals to the store.
  • the pedestrian's location and avatar When the number of arrivals of the target pedestrian is greater than the preset threshold of the number of arrivals, the pedestrian's location and avatar will be sent to the store staff to remind and guide the registration.
  • the embodiment of the present disclosure introduces a stranger's face database to remind strangers to come to the store multiple times, guide the store staff to pay attention to potential customers, and improve the operating effect.
  • the web application can be deployed on the edge node, and the edge application service can communicate with the front end (such as the customer visit store business system deployed on the edge end) to obtain the store visit record of the user.
  • the front end can view user records through a browser, or specify categories of user visit records.
  • the web application can also be deployed on the cloud, and the customer in-store business system deployed on the edge can communicate through the cloud application service to obtain the in-store record, and the cloud web page is responsible for displaying it.
  • An embodiment of the present disclosure also provides a business management platform, including a memory; and a processor coupled to the memory, the processor is configured to execute the operation described in any one of the preceding items based on instructions stored in the memory. The steps of the business management method described above.
  • the service management platform may include: a processor 910, a memory 920, a bus system 930, and a transceiver 940, wherein the processor 910, the memory 920, and the transceiver 940 pass through the bus
  • the system 930 is connected, the memory 920 is used to store instructions, and the processor 910 is used to execute the instructions stored in the memory 920 to control the transceiver 940 to send signals.
  • the transceiver 940 may obtain the monitoring information of the edge node through the service delivery platform under the control of the processor 910, and the processor 910 creates one or more transactions according to the obtained monitoring information of the edge node, and the transaction includes a or multiple AI services, AI service orchestration scripts, image acquisition devices, and available edge nodes, the transceiver 940 sends the created transaction to one or more available edge nodes through the service delivery platform under the control of the processor 910 .
  • the processor 910 can be a central processing unit (Central Processing Unit, CPU), and the processor 910 can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the memory 920 may include read-only memory and random-access memory, and provides instructions and data to the processor 910 .
  • a portion of memory 920 may also include non-volatile random access memory.
  • memory 920 may also store device type information.
  • bus system 930 may also include a power bus, a control bus, and a status signal bus.
  • bus system 930 may also include a power bus, a control bus, and a status signal bus.
  • various buses are labeled as bus system 930 in FIG. 9 for clarity of illustration.
  • the processing performed by the processing device may be completed by an integrated logic circuit of hardware in the processor 910 or instructions in the form of software. That is, the method steps in the embodiments of the present disclosure may be implemented by a hardware processor, or by a combination of hardware and software modules in the processor.
  • the software module may be located in storage media such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers, and the like.
  • the storage medium is located in the memory 920, and the processor 910 reads the information in the memory 920, and completes the steps of the above method in combination with its hardware. To avoid repetition, no detailed description is given here.
  • An embodiment of the present disclosure also provides a computer storage medium, the computer storage medium stores executable instructions, and when the executable instructions are executed by a processor, the service management method provided in any of the above-mentioned embodiments of the present disclosure can be implemented.
  • the service management The method can obtain the monitoring information of the edge node through the service delivery platform; according to the acquired monitoring information of the edge node, create one or more transactions, the transaction includes one or more AI services, AI service orchestration scripts, image acquisition equipment, available Edge nodes; through the service delivery platform, the created transactions are sent to one or more available edge nodes, thereby realizing a cloud-native edge node management method, the transaction delivery method is efficient and convenient, and the transaction management mechanism is efficient and flexible.
  • the method for driving the service management of the service management platform by executing executable instructions is basically the same as the service management method provided in the above-mentioned embodiments of the present disclosure, and details are not repeated here.
  • the business management method, platform, service delivery system, and computer storage medium provided by the embodiments of the present disclosure implement a cloud-native edge through the business management platform scheduling service delivery platform for transaction delivery, container image construction, and edge node monitoring.
  • the node management method, the efficient and convenient transaction delivery method, the efficient and flexible transaction management mechanism, and the flexible construction, convenience and controllability of transactions are guaranteed through AI service orchestration.
  • the functional modules/units in the system, and the device can be implemented as software, firmware, hardware, and an appropriate combination thereof.
  • the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be composed of several physical components. Components cooperate to execute.
  • Some or all of the components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit.
  • Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media).
  • computer storage media includes both volatile and nonvolatile media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. permanent, removable and non-removable media.
  • Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, tape, magnetic disk storage or other magnetic storage devices, or can Any other medium used to store desired information and which can be accessed by a computer.
  • communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media .

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种业务管理方法、平台及服务交付系统、计算机存储介质，该业务管理方法包括：通过服务交付平台获取边缘节点的监控信息，从获取的监控信息中选择可用边缘节点；创建一个或多个事务，每个所述事务包括一个或多个人工智能AI服务；通过所述服务交付平台将创建的事务下发至所述可用边缘节点。本公开实施例实现了一种云原生的边缘节点管理方法，事务下发方式高效便捷，事务管理机制高效灵活，通过AI服务编排，保证事务的灵活构建与便捷可控。

Description

业务管理方法、平台及服务交付系统、计算机存储介质 技术领域
本公开实施例涉及但不限于智能系统技术领域，尤其涉及一种业务管理方法、平台及服务交付系统。
背景技术
边缘计算，是指在靠近物或数据源头的一侧，采用网络、计算、存储、应用核心能力为一体的边缘设备平台，就近提供最近端服务。其应用程序在边缘侧发起，产生更快的网络服务响应，满足行业在实时业务、应用智能、安全与隐私保护等方面的基本需求。云端计算，可以实时接收或访问边缘计算的历史数据。
随着信息技术的发展，边缘设备上部署的人工智能(Artificial Intelligence，AI)服务越来越多，边缘设备数量不断增加，如何高效管理交付这些业务，成为关键问题。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
本公开实施例提供了一种业务管理方法，包括：通过服务交付平台获取边缘节点的监控信息，从获取的监控信息中选择可用边缘节点；创建一个或多个事务，每个创建的所述事务包括一个或多个人工智能AI服务；通过所述服务交付平台将创建的事务下发至所述可用边缘节点。
在示例性实施例中，每个创建的所述事务还包括以下至少之一：AI服务编排脚本、所述事务对应的图像采集设备、所述事务对应的可用边缘节点。
在示例性实施例中，所述方法还包括：对所述可用边缘节点进行初始化。
在示例性实施例中，所述对所述可用边缘节点进行初始化，包括：
选择应用，所述应用为定义所述边缘节点运行的控制管理程序；
从所述服务交付平台获取镜像列表，从所述镜像列表中选择基础容器镜像；
根据选择的所述应用和基础容器镜像，通过所述服务交付平台构建应用容器镜像并将所述应用容器镜像存储至容器仓库，通过所述服务交付平台控制所述可用边缘节点下载应用容器镜像并构建和启动应用容器。
在示例性实施例中，所述应用包括：接收下发事务、动态加载事务、解析服务编排事务和启动线程运行事务。
在示例性实施例中，所述创建一个事务，包括：
选择一个或多个AI服务；
根据选择的一个或多个AI服务，生成AI服务编排脚本；
选择一个或多个图像采集设备；
将所述一个或多个AI服务、AI服务编排脚本、选择的图像采集设备与可用边缘节点信息创建为一个事务。
在示例性实施例中，所述方法还包括：
通过所述服务交付平台获取所述可用边缘节点的事务下发或更新状态。
在示例性实施例中，所述方法还包括：
通过所述服务交付平台对所述边缘节点上的应用进行控制，所述控制包括以下至少之一：创建、启动、停止和更新。
在示例性实施例中，所述方法还包括：
建立所述边缘节点的应用和事务的关联关系，所述应用为定义所述边缘节点运行的控制管理程序；
在多个所述边缘节点挂载统一的共享存储，所述共享存储被配置为存储多个所述边缘节点的应用和事务配置数据。
在示例性实施例中，当所述可用边缘节点发生故障时，所述方法还包括：
检测是否存在第一边缘节点，所述第一边缘节点为空闲的边缘节点，当存在所述第一边缘节点时，将所述发生故障的边缘节点的应用和事务迁移至所述第一边缘节点；
当不存在所述第一边缘节点时，检测是否存在第二边缘节点，所述第二边缘节点的应用版本与所述发生故障的边缘节点的应用版本一致且所述第二边缘节点存在富余的可以接收所述发生故障的边缘节点的事务的资源，当存在所述第二边缘节点时，删除所述发生故障的边缘节点的应用和事务的关联关系，将所述发生故障的边缘节点的事务迁移至所述第二边缘节点；
当不存在所述第一边缘节点和所述第二边缘节点时，产生告警信息。
在示例性实施例中，所述方法还包括：
获取一个或多个所述可用边缘节点的AI服务处理结果，并展示。
在示例性实施例中，通过以下任意一种或多种服务通道：消息队列遥测传输协议、远程字典服务、分布式发布订阅消息系统、网络套接字，获取所述一个或多个可用边缘节点的AI服务处理结果。
在示例性实施例中，一个所述边缘节点并行运行多个事务，每个所述事务启动为一个线程。
本公开实施例还提供了一种业务管理平台，包括存储器；和耦接至所述存储器的处理器，所述处理器被配置为基于存储在所述存储器中的指令，执行如上任一项所述的业务管理方法的步骤。
本公开实施例还提供了一种服务交付系统，包括如上所述的业务管理平台，还包括服务交付平台以及一个或多个边缘节点，所述业务管理平台、服务交付平台和边缘节点通过网络相互连接；
所述服务交付平台被配置为，对所述一个或多个边缘节点进行监控与控制，并将所述业务管理平台创建的事务下发至所述一个或多个边缘节点；
所述边缘节点被配置为，接收所述服务交付平台下发的事务，并对所述事务进行处理，并返回处理结果至所述业务管理平台
在示例性实施例中，该服务交付系统还包括一个或多个图像采集设备，所述边缘节点和图像采集设备通过网络相互连接；
所述边缘节点对所述事务进行处理，包括：根据接收的事务，获取对应的一个多个所述图像采集设备采集的图像或视频数据，并对所述图像或视频数据进行处理。
在示例性实施例中，所述边缘节点并行运行多个事务，所述事务包括客户人脸注册事务、客户识别事务、到店频次分析事务和陌生人提醒事务，其中：
客户人脸注册事务包括人脸检测服务、人脸缩放服务、人脸关键点检测服务、人脸对齐服务、人脸特征提取服务和人脸特征存储服务；
客户识别事务包括视频解码服务、图像缩放服务、目标检测服务、目标跟踪服务、人脸检测服务、人脸缩放服务、人脸质量评价服务、人脸对齐服务、人脸属性判断服务、人脸特征提取服务、人脸检索服务和陌生人注册服务；
到店频次分析事务包括记录上传服务和记录统计服务；
陌生人提醒事务包括陌生人到店情况统计服务和提醒服务。
本公开实施例还提供了一种计算机存储介质，其上存储有计算机程序，该程序被处理器执行时实现如上任一项所述的业务管理方法。
在阅读理解了附图和详细描述后，可以明白其他方面。
附图说明
附图用来提供对本公开技术方案的进一步理解，并且构成说明书的一部分，与本公开的实施例一起用于解释本公开的技术方案，并不构成对本公开的技术方案的限制。附图中各部件的形状和大小不反映真实比例，目的只是示意说明本公开内容。
图1为本公开示例性实施例一种业务管理方法的流程示意图；
图2为本公开示例性实施例一种边缘节点初始化的流程示意图；
图3为本公开示例性实施例一种事务下发/更新的流程示意图；
图4为本公开示例性实施例一种车辆违章检测事务中多个AI服务联动的示意图；
图5a和图5b为本公开示例性实施例两种服务交付系统的结构示意图；
图6为本公开示例性实施例一种到店频次检测事务应用场景示意图；
图7为本公开示例性实施例一种到店频次检测事务中客户识别流程示意图；
图8为本公开示例性实施例一种到店频次检测事务中客户到店频次检测流程示意图；
图9为本公开示例性实施例一种业务管理平台的结构示意图。
具体实施方式
为使本公开的目的、技术方案和优点更加清楚明白，下文中将结合附图对本公开的实施例进行详细说明。注意，实施方式可以以多个不同形式来实施。所属技术领域的普通技术人员可以很容易地理解一个事实，就是方式和内容可以在不脱离本公开的宗旨及其范围的条件下被变换为各种各样的形式。因此，本公开不应该被解释为仅限定在下面的实施方式所记载的内容中。在不冲突的情况下，本公开中的实施例及实施例中的特征可以相互任意组合。
除非另外定义，本公开实施例公开使用的技术术语或者科学术语应当为本公开所属领域内具有一般技能的人士所理解的通常意义。本公开实施例中使用的“第一”、“第二”以及类似的词语并不表示任何顺序、数量或者重要性，而只是用来区分不同的组成部分。“包括”或者“包含”等类似的词语意指出该词前面的元件或物件涵盖出现在该词后面列举的元件或者物件及其等同，而不排除其他元件或者物件。
如图1所示，本公开实施例提供了一种业务管理方法，包括如下步骤：
步骤101、通过服务交付平台获取边缘节点的监控信息，从获取的监控信息中选择可用边缘节点；
本公开实施例的业务管理方法应用在业务管理平台上，业务管理平台为服务发起端，边缘节点为事务的运行单元，服务交付平台通过云网络对边缘节点进行控制，在一些示例性实施方式中，边缘节点在交付时预安装边缘组件(Edge part)，成为K8S节点(Node)。边缘节点的核心应用(APP)可以通过Kubernetes下发。
Kubernetes(简称K8S)，是一个全新的基于容器技术的分布式架构解决方案，是一个开源的容器集群管理系统。一个Kubernetes集群一般包含一个控制(Master)节点和多个Node节点。Master节点是K8S的集群控制节点，每个K8S集群里需要有一个Master节点来负责整个集群的管理和控制，基本上K8S所有的控制命令都是发给它，它来负责实际的执行过程。Node节点是K8S集群中的工作负载节点，每个Node节点都会被Master节点分配一些工作负载，当某个Node节点宕机时，其上的工作负载会被Master节点自动转移到其它Node节点上去。
在一些示例性实施方式中，所述方法还包括：对可用边缘节点进行初始化。
在一些示例性实施方式中，对可用边缘节点进行初始化，包括：
选择应用(APP)，所述应用为定义边缘节点运行的控制管理程序；
从服务交付平台获取镜像列表，从镜像列表中选择基础容器镜像；
根据选择的应用和基础容器镜像，通过服务交付平台构建应用容器镜像，并将构建的应用容器镜像存储至容器仓库，通过服务交付平台控制可用边缘节点下载应用容器镜像并构建和启动应用容器。
如图2所示，业务管理平台根据当前需要的应用的版本，选择该版本的应用；业务管理平台从服务交付平台获取基础容器镜像列表，从基础容器镜像列表中选择基础容器镜像；业务管理平台从服务交付平台获取边缘节点的监控信息，从获取到的边缘节点的监控信息中选择可用边缘节点；业务管理平台根据选择的应用以及基础容器镜像，通过服务交付平台构建应用容器镜像并将所述应用容器镜像存储至容器仓库，通过服务交付平台控制可用边缘节点下载应用容器镜像并构建和启动应用容器。
本公开实施例中，基础容器镜像可以为红帽子(Redhat)、乌班图(Ubuntu)等Linux操作系统，基础容器镜像可以提供应用运行的基础环境。
在一些示例性实施方式中，应用可以包括：接收下发事务、动态加载事务、解析服务编排和启动线程运行事务等。
在一些示例性实施方式中，应用还可以包括：视频解码事务和数据栈管理事务等。当边缘节点需要视频解码服务时，应用可以包括视频解码事务和数据栈管理事务，其中，视频解码事务负责对接收到的视频进行解码，数据栈管理事务负责存储解码后的图像数据。在另一些示例性实施方式中，当边缘节点通过网络直接接收图像数据时，应用可以不用包括视频解码事务和数据栈管理事务。
本实施例中，对可用边缘节点进行初始化后，可用边缘节点启动了上述应用，即具备了接收下发事务、动态加载事务、解析服务编排和启动线程运行事务等的能力。
步骤102、创建一个或多个事务，每个创建的事务包括一个或多个AI服务；
在一些示例性实施方式中，每个创建的事务还包括以下至少之一：AI服务编排脚本、该事务对应的图像采集设备、该事务对应的可用边缘节点。
在一些示例性实施方式中，创建一个事务，包括：
选择一个或多个AI服务；
根据选择的一个或多个AI服务生成AI服务编排脚本；
选择一个或多个图像采集设备；
将一个或多个AI服务的代码库、AI服务编排脚本、选择的图像采集设备与可用边缘节点信息创建为一个事务。
如图3所示，在事务下发及事务启动时，业务管理平台选择一个或多个AI服务，并根据选择的一个或多个AI服务生成AI服务编排脚本；业务管理平台选择一个或多个图像采集设备；业务管理平台从服务交付平台获取边缘节点信息，从获取的边缘节点信息中选择可用边缘节点；业务管理平台将一个或多个AI服务代码库与AI服务编排脚本打包，并下发至可用边缘节点；业务管理平台通过服务交付平台获取边缘节点的事务下发或更新状态。当边缘节点接收到下发的事务时，下载事务包，动态加载AI库，解析服务编排脚本，创建AI服务调用次序，然后启动事务线程。
在一些示例性实施方式中，一个边缘节点并行运行多个事务，每个事务启动为一个线程。
随着语音识别、图像识别等AI技术的发展，越来越多的AI服务投入使用，但单一AI算法模型，往往只针对单一问题，在某些场景一个事务往往需要多个AI服务联动，如图4所示，以违章停车监测为例，如车辆图片输入车型识别模型，可以识别出车型，如需识别车型则需要输入不同AI服务。但是应用场景往往需要通过摄像头等数据源解决如违章停车监测等场景。监控公共场所的车辆停放情况，判断核心区域是否有违章停车，并进而识别违章车辆的车牌号码、品牌型号，通过车型识别算法服务与车牌识别算法服务联动检测车辆，实施对应违章处罚。
在一些示例性实施方式中，可以通过网页(web)页面或其它可视化的方式绘制AI服务编排流程，根据流程图生成对应的AI服务编排脚本，实现多个AI服务联动。
步骤103、通过服务交付平台将创建的事务下发至该事务对应的一个或多个可用边缘节点。
在一些示例性实施方式中，所述方法还包括：
通过服务交付平台获取可用边缘节点的事务下发或更新状态。
如图3所示，边缘节点接收到下发的事务后，先下载事务包，再动态加载AI库，解析服务编排脚本并创建AI服务调用次序，然后启动事务线程，事务启动成功后，更新边缘节点状态。
在一些示例性实施方式中，所述方法还包括：
通过服务交付平台对边缘节点上的应用进行控制，控制包括以下至少之一：创建、启动、停止和更新等。
示例性的，服务交付平台可基于K8S KubeEdge来设计开发，边缘节点安装Edge part来管理运行APP容器，服务交付平台通过Edge part来管理APP容器的生命周期(创建、启动、停止、更新等)。本实施例中，节点控制依赖于K8S对容器的控制能力实现，所以要将其放到服务交付平台，如果 将其放到业务管理平台，即由业务管理平台直接控制边缘节点，那就需要定制自己的控制(创建、更新、启动、停止)逻辑及协议，那样会增大开发工作量。
在一些示例性实施方式中，所述方法还包括：
获取一个或多个可用边缘节点的AI服务处理结果，并展示。
在一些示例性实施方式中，通过以下任意一种或多种服务通道：消息队列遥测传输协议(Message Queuing Telemetry Transport，MQTT)、远程字典服务(Remote Dictionary Server，Redis)、分布式发布订阅消息系统(如Kafka)、网络套接字(WebSocket，一种基于传输控制协议(Transmission Control Protocol，TCP)的全双工通信协议)等，获取一个或多个可用边缘节点的AI服务处理结果。
本公开实施例提供的业务管理方法，通过业务管理平台调度服务交付平台进行事务下发、容器镜像构建、边缘节点监控等，实现了一种云原生的边缘节点管理方法，事务下发方式高效便捷，事务管理机制高效灵活，通过AI服务编排，保证事务的灵活构建与便捷可控。
本公开实施例还提供了一种服务交付系统，包括如前所述的业务管理平台，还包括服务交付平台以及一个或多个边缘节点。所述业务管理平台、服务交付平台和边缘节点通过网络相互连接，所述边缘节点和图像采集设备通过网络相互连接。
服务交付平台被配置为，对一个或多个边缘节点进行监控与控制，并将业务管理平台创建的事务下发至一个或多个边缘节点；
边缘节点被配置为，接收服务交付平台下发的事务，并对事务进行处理，并返回处理结果至业务管理平台。
在一些示例性实施方式中，该服务交付系统还包括一个或多个图像采集设备，边缘节点和图像采集设备之间通过网络相互连接；
边缘节点对事务进行处理，包括：根据接收的事务，获取对应的一个多个图像采集设备采集的图像或视频数据，并对图像或视频数据进行处理。
如图5a和图5b所示，本公开实施例提供的服务交付系统，至少包括以下组成部分：业务管理平台、服务交付平台和边缘节点。
(1)业务管理平台
业务管理平台为服务发起端，其功能模块包括但不限于边缘节点管理模块、摄像头管理模块、事务管理模块、APP管理模块、AI服务管理模块、AI服务展示模块等。
边缘节点管理模块：负责配置业务管理平台的全部边缘节点信息，与服务交互平台的边缘监控模块交互获取当前边缘节点的资源使用情况。
摄像头管理模块：负责记录所有摄像头视频流的配置信息(示例性的，配置信息包括流地址、摄像头型号、摄像头位置、摄像头厂家等信息)，以及每个摄像头与事务的关联关系(即事务的数据源来源于哪个摄像头或哪几个摄像头)。
事务管理模块：将一种AI服务能力定义为一种事务，创建事务需配置AI服务、生成AI服务编排脚本、选择摄像头、选择可用边缘节点等，图5中以到店频次或离岗检测为例定义一个事务，该事务需要与其他模块交互以获取必要的配置信息。
APP管理模块：APP指的是定义边缘节点运行的控制管理程序，本公开实施例中，APP主要功能为视频流解码、AI服务管理调度等，APP管理包括启动APP、停止APP、更新APP等。
AI服务管理模块：本模块负责将AI服务打包成可被APP调度的模块，并且定义多个AI服务之间的联动编排关系。本公开实施例中，如何定义AI服务之间的联动编排，需要根据实际的使用场景来定义，本公开对此不作限制。
AI服务展示模块：负责展示来自于边缘节点的AI服务处理结果。
(2)服务交付平台
服务交付平台可基于K8S KubeEdge来设计开发，边缘端安装Edge part来管理运行APP的边缘容器，服务交付平台通过Edge part来管理APP容器的生命周期(创建、启动、停止、删除等)，并通过服务交付平台的Open API， 由业务管理平台调度服务交付平台进行APP容器镜像构建、事务更新、边缘节点控制、以及将边缘节点监控信息返回到业务管理平台。
KubeEdge为Kubernetes的原生边缘计算平台，KubeEdge架构包括两部分，分别是云端和边缘侧，云端负责应用和配置的下发，边缘侧负责运行边缘应用和管理接入设备。
(3)边缘节点
边缘节点为事务的运行单元，边缘节点在交付时预安装Edge part，成为K8S Node节点。边缘节点的核心应用APP通过K8S下发。
APP应用主要包括视频解码事务、数据栈管理事务、接收下发事务、动态加载事务、解析服务编排事务、启动线程运行事务等。
一个边缘节点可以并行运行多个事务，每个事务可以启动为一个线程，不同的事务可以并行执行。
在一些示例性实施例中，一个边缘节点运行两个事务：事务1和事务2，示例性的，事务1可以用于对人脸进行检测；示例性的，事务2可以用于对人体进行检测，由于事务1和事务2对摄像头的拍摄精度和拍摄范围要求可能不同，分开检测有利于对对应的摄像头的拍摄精度和拍摄范围分别进行控制，但是，在回传结果时，可以对同一检测目标的事务1和事务2的检测结果打包回传至业务管理平台。
事务产生的结果由边缘节点统一通过以下任意一种或多种服务通道：消息队列遥测传输协议、远程字典服务、分布式发布订阅消息系统、网络套接字等，上报到业务管理平台。
在一些示例性实施例中，边缘AI事务输出的结果可以包括：展示、告警、通知等信息。
以本方案离岗检测为例，当边缘节点的AI服务处理结果发现当前岗位没有相关人员，就发送如(某位置：离岗)的信息到业务管理平台的AI服务展示模块，AI服务展示模块可以弹出告警框“某位置某人离岗”，并发送告警到相关管理人员。此外，可以通过WebSocket协议等方式，拉取AI服务处理后的画面，显示不在岗人员信息及位置信息。
以本方案到店频次检测为例，当某客户当天第一次出现，AI服务处理结果返回该客户ID及到店时间等信息到AI服务展示模块，AI服务展示模块接收到该客户的信息后，先将该客户的本次到店信息记录到数据库，然后统计该客户指定时间段(如1年)的到店次数，可以在网页(Web)端弹出1年到店次数>预设到店次数阈值(如5次)的提示信息，以便店员可以重点关注该客户。实际使用时，预设到店次数阈值的大小可以根据需要进行调整，如当需要关注陌生客户时，可以弹出初次到店的客户信息，例如：到店次数＝1，预设到店次数阈值的大小可以通过AI服务展示模块进行配置。
当服务交付平台监控到某一边缘节点(如边缘节点1)发生故障(如宕机)时，将故障信息上报到业务管理平台的边缘节点管理模块进行处理，由于本公开实施例中，可用边缘节点同时运行应用(APP)容器与下发的事务，因此可分成以下两种情况实现高可用：
1)当存在空闲的边缘节点(如边缘节点2)时，由于应用打包成应用容器(docker)镜像，因此，应用可以通过K8S高可用能力迁移的能力，迁移到其它可用空闲边缘节点(如边缘节点2)。但是，事务数据无法直接从发生故障的边缘节点1迁移到空闲的边缘节点2，因此，在一些示例性实施例中，可以在业务管理平台的事务管理模块中，将每个边缘节点的应用和事务关联，并在多个边缘节点挂载统一的共享存储，该共享存储用于存储多个边缘节点的应用事务配置数据。因此，当将边缘节点1的应用和事务数据迁移到边缘节点2时，只需在边缘节点2启动该应用时读取边缘节点1的应用事务配置数据即可。
示例性的，边缘节点的应用和事务关联关系可以表示为APP-ID1(事务1，事务2)，其中，APP-ID1的ID1可以是该应用打包成的docker镜像的名字，通过这个ID关联起这个应用与事务。
当不存在空闲的边缘节点，但是有边缘节点的应用版本与发生故障的边缘节点的应用版本一致，同时该边缘节点的资源使用情况满足接受更多事务时，那么就在业务管理平台上，通过边缘节点管理模块删除事务管理模块中发生故障的边缘节点的应用和事务的关联关系，例如，删除边缘节点1上运行的事务1、2的记录信息与APP-ID1的关联关系，并在对应的共享存储上 删除APP-ID1与事务1、2的关联关系，以避免边缘节点1修复时，一个事务多处运行，通过事务管理模块将对应事务1、2下发到起其它可用同版本边缘节点(如边缘节点2)上运行，更新共享存储中边缘节点应用和事务的关联关系，如配置APP-ID2(事务4，事务1，事务2)，更新事务管理模块中边缘节点应用和事务的关联关系，如记录APP-ID2(事务4，事务1，事务2)。
2)当即无空闲边缘节点也无满足需要的运行同版本APP的边缘节点时，由边缘节点管理模块触发报警机制，发短信、微信、邮件等触发人工干预处理。
当某一摄像头发生故障时，边缘节点的应用执行视频编解码时，发现程序无法正常运行，发送告警信息，服务交付平台的边缘监控模块收到信息后，边缘监控模块将接收到的信息反馈给业务管理平台的边缘节点管理模块，触发报警机制，发短信、微信、邮件等触发人工干预处理。本实施例中，边缘监控模块放到服务交付平台是因为可以直接使用K8S的监控系统如Prometheus(一个开源的系统监控和报警系统)，业务管理平台更多的是侧重于业务能力管理，而服务交付平台侧重的是交付运维管理。
当某个事务需要对AI服务进行优化，本公开的业务管理方法可以完成事务热更新。以更新事务2为例，首先通过业务管理平台的事务管理模块，调用服务交付平台的事务更新模块，通知边缘节点1的应用APP-ID1删除事务2，APP-ID1收到删除命令，通过主线程停止退出事务2线程，集中共享存储上本节点边缘事务配置数据APP-ID1(事务1，事务2)删除，变为APP-ID1(事务1)，该边缘节点向服务交付平台的事务更新模块上报本次删除任务已完成，然后业务管理平台的事务管理模块重新生成事务下发，下发的边缘节点不用选择，依然为边缘节点1，启动事务下发及事务启动流程，并更新事务管理模块中边缘节点1的应用和事务关联关系，以及共享存储中边缘节点1的事务配置数据APP-ID1(事务1，事务2)。
下面以到店频次检测事务来举例说明本公开实施例的业务管理方法。
商业、店铺等经营中，往往需要掌握客户到店的情况，包括每天，每周， 每月等到店统计，来对客户进行画像，调整经营策略。之前往往通过店员对目标客户到访进行人工记录的方式来获得数据，但是这种方式低效而且数据获取不准确，容易出现漏报的情况。针对这种情况，本公开实施例设计基于一种边缘计算的客户到店频次统计系统，可提供一种安全、高效、智能的客户到店统计方式。该系统在硬件上包括摄像头、网络、边缘设备(即边缘节点)，在软件上包括视频流处理模块、目标检测模块、目标跟踪模块、人脸矫正模块，人脸识别模块等，以及数据存储、分析计算、展示以及同步云端等功能模块，部署架构如图6所示。
本公开实施例通过将边缘设备部署到店铺内或者靠近店铺，边缘设备靠近应用场景，可实现低延时的视频传输，高实时可见的客户到店记录或提醒。
本系统由软件硬件两个部分组成：
硬件包括：图像采集设备(如摄像头)，图像处理设备(如边缘节点)，能连接图像采集设备与图像处理设备的网络设备，服务交付平台，业务管理平台。
软件包括：本系统的软件主要运行在边缘节点上，包括视频解码模块、目标检测模块、目标跟踪模块、人脸矫正模块、人脸识别模块等。本公开所说的“模块”，也可以称为“服务”。基于每个模块的处理结果对客户到店情况进行存储，按需进行每天、每周、每月、每年等到店频次分析展示。
目标检测模块被配置为，从一幅场景(图片)中找出目标，包括检测(where)和识别(what)两个过程。
目标跟踪模块被配置为，在连续的视频序列中，建立所要跟踪物体的位置关系，得到物体完整的运动轨迹。目标跟踪模块根据给定的图像第一帧的目标坐标位置，计算在下一帧图像中目标的确切位置。在运动的过程中，目标可能会呈现一些图像上的变化，比如姿态或形状的变化、尺度的变化、背景遮挡或光线亮度的变化等。目标跟踪技术是计算机视觉研究领域的热点之一，并得到广泛应用。相机的跟踪对焦、无人机的自动目标跟踪等都需要用到了目标跟踪技术。另外还有特定物体的跟踪，比如人体跟踪，交通监控系统中的车辆跟踪，人脸跟踪和智能交互系统中的手势跟踪等。
人脸矫正模块被配置为，在检测到的人脸角度不是很正时，通过人脸关键点检测，以及基于关键点旋转变换等方法使其对齐。
人脸识别模块包括人脸识别子模块和人脸验证子模块，人脸识别子模块被配置为，将一个人脸分类为一个特定的标识(identification)；人脸验证子模块被配置为，确定一对图片是否属于同一人(Verification)。
在一些示例性实施方式中，一个边缘节点上可以并行运行多个事务，示例性的，多个事务包括客户人脸注册事务、客户识别事务、到店频次分析事务和陌生人提醒事务，其中：
客户人脸注册事务包括人脸检测服务、人脸缩放服务、人脸关键点检测服务、人脸对齐服务、人脸特征提取服务和人脸特征存储服务；
客户识别事务包括视频解码服务、图像缩放服务、目标检测服务、目标跟踪服务、人脸检测服务、人脸缩放服务、人脸质量评价服务、人脸对齐服务、人脸属性判断服务、人脸特征提取服务、人脸检索服务和陌生人注册服务；
到店频次分析事务包括记录上传服务和记录统计服务；
陌生人提醒事务包括陌生人到店情况统计服务和提醒服务。
在一些示例性实施方式中，本系统包括但不限于如下功能：
(1)客户人脸注册。
如图7所示，本系统为了识别出客户身份，需要首先基于客户图片完成人脸注册，管理人员可通过边缘节点提供的接口如网页(Web)服务上传图片，图片上传后触发客户注册功能，先将图片传给目标检测模块(该模块包含人脸检测算法)，识别出客户的人脸信息。
某些情况人脸可能出现歪头等情况影响客户识别准确性这时就需要通过人脸矫正模块将检测出的人脸进行人脸对齐，对齐流程包括首先对人脸尺寸进行缩放，调用人脸关键点检测算法识别出人脸关键点信息，传入人脸对齐算法进行人脸对齐，实现歪头矫正。然后调用人脸识别模块提取人脸特征，最后调用人脸存储模块将提取的特征存储到边缘节点中，为保护客户隐私， 提取人脸特征后可删除客户图像。本公开实施例通过将用户数据存储到边缘设备，只存储客户人脸特征，保障客户数据安全和隐私。
(2)客户识别，到店频次分析
本公开实施例以3个摄像头为例来采集，避免一个摄像头因为遮挡等原因影响用户人脸数据提取。如图8所示，先通过视频解码模块对输入视频流进行解码，将解码后图像进行缩放调用目标检测模块检测出图像中行人信息，通过目标跟踪模块判断该行人是否是第一次出现。
当目标跟踪模块判断该行人是第一次出现时，对该行人进行人脸检测，检测出人脸后进行图像缩放，输入人脸质量评价模块判断人脸是否模糊或者遮挡，当人脸质量不合格时，则返回进行下一轮视频流处理；当人脸质量合格时，进行人脸对齐矫正歪头等情况，将矫正后结果输入人脸属性模块，获取年龄、性别等辅助信息，然后调用人脸特征提取模块进行人脸特征提取，用提取到的特征到人脸特征库中检索到匹配的客户，与人脸属性特征进行综合匹配后确定最终客户，将该行人与该客户绑定，记录客户到店时间与客户的轨迹信息。
当目标跟踪模块判断该行人不是第一次出现时，判断行人是否与客户绑定，当行人与客户绑定时，记录客户到店时间与客户的轨迹信息，否则作为陌生人，记录客户到店时间，然后返回进行下一轮视频流处理。
当目标检测后发现目标跟踪的行人已经消失时，记录客户离店时间。本公开实施例通过多个摄像头、目标跟踪、准确人脸识别等技术可更准确的抓取到客户到店的实时信息记录客户到店离店情况，提供给经营者参考。
客户到店记录、店内轨迹信息可以存储在边缘节点上，这样可以保障敏感数据安全，但设备损坏等问题可能会导致数据丢失，因此，可以在保障数据安全的情况下(如加密传输)将客户到店记录、店内轨迹信息上传至云端。
当经营者需要对客户行为进行分析时，提取存储到数据库等存储方式中客户到店、离店时间戳，进行时间戳计算，生成每天到店几次，每周到店几次，每月到店次数统计。某天检测到一次到店记录、一次离店记录视为一次到店。此外，在记录到店记录信息的同时，本公开还记录了客户店内轨迹信 息，可按需对客户在店铺轨迹进行分析，为某些经营者提供客户画像的参考。
(3)陌生人到店统计提醒
如图8所示，对注册的客户可以实现完整客户到店频次统计，以及活动轨迹分析。很多情况会有一些客户没有注册但是也频繁访问店铺，需要提醒店员来关注引导用户注册，提升客户粘性。
本公开针对陌生人，设计了陌生人人脸库，当检索到目标行人非注册客户时，到陌生人人脸库进行检索，当目标行人为非注册陌生人时，先注册再记录到店时间与离店时间。记录完成触发陌生人人到店情况统计，当目标行人的到店次数>预设的到店次数阈值时，发送该行人位置与头像给店员提醒，引导注册。
本公开实施例通过引入陌生人脸库，实现陌生人多次到店提醒，引导店员关注潜在客户，提升经营效果。
(4)分析结果展示
本公开实施例中，可以在边缘节点上部署Web应用，通过边缘应用服务与前端(如边缘端部署的客户到店业务系统)进行通信，获取用户到店记录。可以由前端通过浏览器查看用户记录，或者指定类别用户到店记录。
本公开实施例中，也可以在云端部署Web应用，通过云端应用服务与边缘端部署的客户到店业务系统进行通信获取到店记录，由云端Web页面负责展示。
本公开实施例还提供了一种业务管理平台，包括存储器；和耦接至所述存储器的处理器，所述处理器被配置为基于存储在所述存储器中的指令，执行如前任一项所述的业务管理方法的步骤。
如图9所示，在一个示例中，业务管理平台可包括：处理器910、存储器920、总线系统930和收发器940，其中，该处理器910、该存储器920和该收发器940通过该总线系统930相连，该存储器920用于存储指令，该处理器910用于执行该存储器920存储的指令，以控制该收发器940发送信号。可选地，收发器940可在处理器910的控制下通过服务交付平台获取边缘节点的监控信息，处理器910根据获取的边缘节点的监控信息，创建一个或多 个事务，所述事务包括一个或多个AI服务、AI服务编排脚本、图像采集设备、可用边缘节点，收发器940在处理器910的控制下通过服务交付平台将创建的事务下发至一个或多个可用边缘节点。
应理解，处理器910可以是中央处理单元(Central Processing Unit，CPU)，处理器910还可以是其他通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现成可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
存储器920可以包括只读存储器和随机存取存储器，并向处理器910提供指令和数据。存储器920的一部分还可以包括非易失性随机存取存储器。例如，存储器920还可以存储设备类型的信息。
总线系统930除包括数据总线之外，还可以包括电源总线、控制总线和状态信号总线等。但是为了清楚说明起见，在图9中将各种总线都标为总线系统930。
在实现过程中，处理设备所执行的处理可以通过处理器910中的硬件的集成逻辑电路或者软件形式的指令完成。即本公开实施例的方法步骤可以体现为硬件处理器执行完成，或者用处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等存储介质中。该存储介质位于存储器920，处理器910读取存储器920中的信息，结合其硬件完成上述方法的步骤。为避免重复，这里不再详细描述。
本公开实施例还提供了一种计算机存储介质，该计算机存储介质存储有可执行指令，该可执行指令被处理器执行时可以实现本公开上述任一实施例提供的业务管理方法，该业务管理方法可以通过服务交付平台获取边缘节点的监控信息；根据获取的边缘节点的监控信息，创建一个或多个事务，所述事务包括一个或多个AI服务、AI服务编排脚本、图像采集设备、可用边缘节点；通过服务交付平台将创建的事务下发至一个或多个可用边缘节点，从而实现了一种云原生的边缘节点管理方法，事务下发方式高效便捷，事务管 理机制高效灵活，通过AI服务编排，保证事务的灵活构建与便捷可控。通过执行可执行指令驱动业务管理平台业务管理的方法与本公开上述实施例提供的业务管理方法基本相同，在此不做赘述。
本公开实施例提供的业务管理方法、平台、服务交付系统及计算机存储介质，通过业务管理平台调度服务交付平台进行事务下发、容器镜像构建、边缘节点监控等，实现了一种云原生的边缘节点管理方法，事务下发方式高效便捷，事务管理机制高效灵活，通过AI服务编排，保证事务的灵活构建与便捷可控。
本领域普通技术人员可以理解，上文中所公开方法中的全部或某些步骤、系统、装置中的功能模块/单元可以被实施为软件、固件、硬件及其适当的组合。在硬件实施方式中，在以上描述中提及的功能模块/单元之间的划分不一定对应于物理组件的划分；例如，一个物理组件可以具有多个功能，或者一个功能或步骤可以由若干物理组件合作执行。某些组件或所有组件可以被实施为由处理器，如数字信号处理器或微处理器执行的软件，或者被实施为硬件，或者被实施为集成电路，如专用集成电路。这样的软件可以分布在计算机可读介质上，计算机可读介质可以包括计算机存储介质(或非暂时性介质)和通信介质(或暂时性介质)。如本领域普通技术人员公知的，术语计算机存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储装置、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。此外，本领域普通技术人员公知的是，通信介质通常包含计算机可读指令、数据结构、程序模块或者诸如载波或其他传输机制之类的调制数据信号中的其他数据，并且可包括任何信息递送介质。
虽然本公开所揭露的实施方式如上，但所述的内容仅为便于理解本公开而采用的实施方式，并非用以限定本公开。任何所属领域内的技术人员，在不脱离本公开所揭露的精神和范围的前提下，可以在实施的形式及细节上进行任何的修改与变化，但本公开的专利保护范围，仍须以所附的权利要求书 所界定的范围为准。

Claims (18)

  1. 一种业务管理方法，包括：
    通过服务交付平台获取边缘节点的监控信息，从获取的监控信息中选择可用边缘节点；
    创建一个或多个事务，每个创建的所述事务包括一个或多个人工智能AI服务；
    通过所述服务交付平台将创建的事务下发至所述可用边缘节点。
  2. 根据权利要求1所述的业务管理方法，其中，每个创建的所述事务还包括以下至少之一：AI服务编排脚本、所述事务对应的图像采集设备、所述事务对应的可用边缘节点。
  3. 根据权利要求1所述的业务管理方法，所述方法还包括：对所述可用边缘节点进行初始化。
  4. 根据权利要求3所述的业务管理方法，其中，所述对所述可用边缘节点进行初始化，包括：
    选择应用，所述应用为定义所述边缘节点运行的控制管理程序；
    从所述服务交付平台获取镜像列表，从所述镜像列表中选择基础容器镜像；
    根据选择的所述应用和基础容器镜像，通过所述服务交付平台构建应用容器镜像并将所述应用容器镜像存储至容器仓库，通过所述服务交付平台控制所述可用边缘节点下载应用容器镜像并构建和启动应用容器。
  5. 根据权利要求4所述的业务管理方法，其中，所述应用包括：接收下发事务、动态加载事务、解析服务编排事务和启动线程运行事务。
  6. 根据权利要求1所述的业务管理方法，其中，所述创建一个事务，包括：
    选择一个或多个AI服务；
    根据选择的一个或多个AI服务，生成AI服务编排脚本；
    选择一个或多个图像采集设备；
    将所述一个或多个AI服务、AI服务编排脚本、选择的图像采集设备与可用边缘节点信息创建为一个事务。
  7. 根据权利要求1所述的业务管理方法，所述方法还包括：
    通过所述服务交付平台获取所述可用边缘节点的事务下发或更新状态。
  8. 根据权利要求1所述的业务管理方法，所述方法还包括：
    通过所述服务交付平台对所述边缘节点上的应用进行控制，所述控制包括以下至少之一：创建、启动、停止和更新。
  9. 根据权利要求1所述的业务管理方法，所述方法还包括：
    建立所述边缘节点的应用和事务的关联关系，所述应用为定义所述边缘节点运行的控制管理程序；
    在多个所述边缘节点挂载统一的共享存储，所述共享存储被配置为存储多个所述边缘节点的应用和事务配置数据。
  10. 根据权利要求9所述的业务管理方法，当所述可用边缘节点发生故障时，所述方法还包括：
    检测是否存在第一边缘节点，所述第一边缘节点为空闲的边缘节点，当存在所述第一边缘节点时，将所述发生故障的边缘节点的应用和事务迁移至所述第一边缘节点；
    当不存在所述第一边缘节点时，检测是否存在第二边缘节点，所述第二边缘节点的应用版本与所述发生故障的边缘节点的应用版本一致且所述第二边缘节点存在富余的可以接收所述发生故障的边缘节点的事务的资源，当存在所述第二边缘节点时，删除所述发生故障的边缘节点的应用和事务的关联关系，将所述发生故障的边缘节点的事务迁移至所述第二边缘节点；
    当不存在所述第一边缘节点和所述第二边缘节点时，产生告警信息。
  11. 根据权利要求1所述的业务管理方法，所述方法还包括：
    获取一个或多个所述可用边缘节点的AI服务处理结果，并展示。
  12. 根据权利要求11所述的业务管理方法，其中，
    通过以下任意一种或多种服务通道：消息队列遥测传输协议、远程字典服务、分布式发布订阅消息系统、网络套接字，获取所述一个或多个可用边缘节点的AI服务处理结果。
  13. 根据权利要求1所述的业务管理方法，其中，一个所述边缘节点并行运行多个事务，每个所述事务启动为一个线程。
  14. 一种业务管理平台，包括存储器；和耦接至所述存储器的处理器，所述处理器被配置为基于存储在所述存储器中的指令，执行如权利要求1至13中任一项所述的业务管理方法的步骤。
  15. 一种服务交付系统，包括如权利要求14所述的业务管理平台，还包括服务交付平台以及一个或多个边缘节点，所述业务管理平台、服务交付平台和边缘节点通过网络相互连接；
    所述服务交付平台被配置为，对所述一个或多个边缘节点进行监控与控制，并将所述业务管理平台创建的事务下发至所述一个或多个边缘节点；
    所述边缘节点被配置为，接收所述服务交付平台下发的事务，并对所述事务进行处理，并返回处理结果至所述业务管理平台。
  16. 根据权利要求15所述的服务交付系统，还包括一个或多个图像采集设备，所述边缘节点和图像采集设备通过网络相互连接；
    所述边缘节点对所述事务进行处理，包括：根据接收的事务，获取对应的一个多个所述图像采集设备采集的图像或视频数据，并对所述图像或视频数据进行处理。
  17. 根据权利要求15所述的服务交付系统，其中，所述边缘节点并行运行多个事务，所述事务包括客户人脸注册事务、客户识别事务、到店频次分析事务和陌生人提醒事务；
    客户人脸注册事务包括人脸检测服务、人脸缩放服务、人脸关键点检测服务、人脸对齐服务、人脸特征提取服务和人脸特征存储服务；
    客户识别事务包括视频解码服务、图像缩放服务、目标检测服务、目标跟踪服务、人脸检测服务、人脸缩放服务、人脸质量评价服务、人脸对齐服务、人脸属性判断服务、人脸特征提取服务、人脸检索服务和陌生人注册服务；
    到店频次分析事务包括记录上传服务和记录统计服务；
    陌生人提醒事务包括陌生人到店情况统计服务和提醒服务。
  18. 一种计算机存储介质，其上存储有计算机程序，该程序被处理器执行时实现如权利要求1至13中任一项所述的业务管理方法。
PCT/CN2023/071185 2022-01-25 2023-01-09 业务管理方法、平台及服务交付系统、计算机存储介质 WO2023142986A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/282,802 US20240168825A1 (en) 2022-01-25 2023-01-09 Service Management Method, Platform, Service Delivery System, and Computer Storage Medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210090084.7A CN114490063A (zh) 2022-01-25 2022-01-25 业务管理方法、平台及服务交付系统、计算机存储介质
CN202210090084.7 2022-01-25

Publications (1)

Publication Number Publication Date
WO2023142986A1 true WO2023142986A1 (zh) 2023-08-03

Family

ID=81474523

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/071185 WO2023142986A1 (zh) 2022-01-25 2023-01-09 业务管理方法、平台及服务交付系统、计算机存储介质

Country Status (3)

Country Link
US (1) US20240168825A1 (zh)
CN (1) CN114490063A (zh)
WO (1) WO2023142986A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114490063A (zh) * 2022-01-25 2022-05-13 京东方科技集团股份有限公司 业务管理方法、平台及服务交付系统、计算机存储介质
CN114979246A (zh) * 2022-05-18 2022-08-30 京东方科技集团股份有限公司 业务管理方法、系统、配置服务器及边缘计算设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110191007A (zh) * 2019-06-27 2019-08-30 广州虎牙科技有限公司 节点管理方法、系统及计算机可读存储介质
CN112799789A (zh) * 2021-03-22 2021-05-14 腾讯科技（深圳）有限公司 一种节点集群管理方法、装置、设备及存储介质
CN113381511A (zh) * 2021-06-02 2021-09-10 深圳市莱达四维信息科技有限公司 变电站在线智能巡视系统
WO2021211905A1 (en) * 2020-04-15 2021-10-21 Ramakanth Gunuganti Flow management with services
CN113630372A (zh) * 2020-05-09 2021-11-09 中移物联网有限公司 一种边缘计算的云边协同系统
CN114490063A (zh) * 2022-01-25 2022-05-13 京东方科技集团股份有限公司 业务管理方法、平台及服务交付系统、计算机存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110191007A (zh) * 2019-06-27 2019-08-30 广州虎牙科技有限公司 节点管理方法、系统及计算机可读存储介质
WO2021211905A1 (en) * 2020-04-15 2021-10-21 Ramakanth Gunuganti Flow management with services
CN113630372A (zh) * 2020-05-09 2021-11-09 中移物联网有限公司 一种边缘计算的云边协同系统
CN112799789A (zh) * 2021-03-22 2021-05-14 腾讯科技（深圳）有限公司 一种节点集群管理方法、装置、设备及存储介质
CN113381511A (zh) * 2021-06-02 2021-09-10 深圳市莱达四维信息科技有限公司 变电站在线智能巡视系统
CN114490063A (zh) * 2022-01-25 2022-05-13 京东方科技集团股份有限公司 业务管理方法、平台及服务交付系统、计算机存储介质

Also Published As

Publication number Publication date
US20240168825A1 (en) 2024-05-23
CN114490063A (zh) 2022-05-13

Similar Documents

Publication Publication Date Title
WO2023142986A1 (zh) 业务管理方法、平台及服务交付系统、计算机存储介质
US10410278B2 (en) Method and apparatus for integrated image capture for vehicles to track damage
JP6972119B2 (ja) ストリームデータにおける空間変化検出器
US9892379B1 (en) Monitoring and notification of delivered packages
TW201915885A (zh) 貨物訂單處理方法、裝置、伺服器、購物終端及系統
US11436819B2 (en) Consolidation and history recording of a physical display board using an online task management system
CN105591821B (zh) 监控系统和业务系统
KR102240997B1 (ko) 배송 서비스 시스템
TW202009860A (zh) 任務處理方法、裝置及系統
US20150269549A1 (en) Synchronizing scan activity with loss prevention cameras
US11443259B2 (en) Automatic floor-level retail operation decisions using video analytics
US20180157917A1 (en) Image auditing method and system
RU2730112C1 (ru) Система и способ идентификации объектов в составном объекте
KR20220051129A (ko) 비동기적으로 큐에 더해진 요청의 오류를 검출하는 시스템 및 방법
KR20200112265A (ko) 배송 서비스 시스템
US11853881B2 (en) Image analysis-based classification and visualization of events
CN113233069B (zh) 智能仓储出库控制方法、装置、电子设备、及存储介质
US11600085B2 (en) Automated training data collection for object detection
US10074098B2 (en) Demographic information collection and content display based on demographic information
CN110400104A (zh) 货物运输和订单分配方法、设备及存储介质
US20230141849A1 (en) Workflow management based on recognition of content of documents
CN111722889A (zh) 车型展示方法、终端设备及存储介质
US11599742B2 (en) Dynamic image recognition and training using data center resources and data
RU2785327C1 (ru) Система компьютерного зрения в ритейле
US20240046056A1 (en) Logistics control system and method by using camera

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23745871

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18282802

Country of ref document: US