KR102331347B1 - An extraction-system using dispersion deep learning information analysis management based cloud - Google Patents

Abstract

The present invention relates to an object extraction system using cloud-based distributed deep learning information analysis and management, and by forming a virtual network, performing distributed deep learning using it, and sharing the information with a large number of users, big data It is possible to reduce the cost and minimize the time required for data use by distributing the facilities according to the system.
To this end, the object extraction system using the cloud-based distributed deep learning information analysis management manages the cloud use application and administrator's authority management service, data backup and recovery, security, etc., and analyzes statistics and logs and provides them in the form of a dashboard A hub service unit (1) that provides or documents documents; and functions such as resource creation, resource group creation and user registration in connection with the hub service unit (1), system infrastructure operation and management such as security management, statistics and analysis It provides an object extraction system using cloud-based distributed deep learning information analysis and management, characterized in that it consists of a cloud service unit (2) in charge of a report function based on it.

Description

Object extraction system using cloud-based distributed deep learning information analysis management {AN EXTRACTION-SYSTEM USING DISPERSION DEEP LEARNING INFORMATION ANALYSIS MANAGEMENT BASED CLOUD}

The present invention relates to an object extraction system using cloud-based distributed deep learning information analysis and management, and by forming a virtual network, performing distributed deep learning using it, and sharing the information with a large number of users, big data It is possible to reduce the cost and minimize the time required for data use by distributing the facilities according to the system.

In general, deep learning is an information retrieval and extraction method using a learning method that provides more accurate results by repeatedly executing commands under the same conditions several times to hundreds of thousands of times in order to analyze the information desired by the user.

Such deep learning is increasing the accuracy to meet the requirements of users in the results, but as a side effect, the occurrence of big data and the cost of storing, managing, and operating such big data are excessively generated. do.

For example, a storage for storing data according to deep learning becomes large or a high-capacity storage device is required, resulting in excessive cost.

In addition, when the results obtained through deep learning learning and raw data before deep learning learning are also stored and managed in the main server and main management part, the user to use it is connected to the main server or main management part Therefore, it takes a lot of time for users to search for the information they want and use the data.

Republic of Korea Patent Publication No. 10-2019-0069655 (cloud computing-based deep learning development platform providing method apparatus, published on June 20, 2019) discloses a cloud computing-based deep learning development platform providing method apparatus. Deep learning development platform according to an embodiment

The providing method includes receiving a deep learning-related service request from a front-end module, and performing the requested deep-learning-related service using a backend module, the front-end module and the back-end module publishes a technology implemented in a cloud computing server.

Republic of Korea Patent Publication No. 10-2019-0080665 (a method of providing a cloud-based intelligent remote security control service using deep learning technology, published on July 08, 2019) uses deep learning technology in a cloud environment for intelligent image analysis, remote We publish a method of providing cloud-based intelligent remote security control service using deep learning technology to effectively provide services such as control access control management and IOT monitoring.

Republic of Korea Patent Publication No. 10-2019-0069655 (Cloud computing-based deep learning development platform providing method device, published on June 20, 2019) Republic of Korea Patent Publication No. 10-2019-0080665 (A method of providing a cloud-based intelligent remote security control service using deep learning technology, published on July 08, 2019)

In order to overcome the above problems, the present invention provides a system capable of reducing large-scale facilities and costs required for processing and storing large amounts of data according to object analysis through deep learning through distributed processing and management. There is a purpose.

In addition, it is an object of the present invention to provide a system capable of preventing traffic from occurring in a specific section and acquiring a large amount of object information when a plurality of users access the network at the same time.

In order to achieve the object of the present invention, the object extraction system using cloud-based distributed deep learning information analysis management manages cloud use application and administrator's authority management service, data backup and recovery, security, etc., and analyzes statistics and logs The hub service unit (1) and the hub service unit (1) to provide in the form of a dashboard or to provide documentation and functions such as resource creation, resource group creation and user registration, and system infrastructure operation such as security management and To provide an object extraction system using cloud-based distributed deep learning information analysis management, characterized in that it consists of a cloud service unit (2) responsible for management, statistics, and analysis-based report functions.

In addition, the hub service unit 1 includes a user management unit 11 that performs user subscription and authority management, an infrastructure operation management unit 12 that manages data backup and storage, monitoring and notification functions according to system operation, and It consists of a report management unit 13 that manages history such as backup and security log, user statistics, and security statistics, and the cloud service unit 2 groups according to the type of object, information about the user, and ID for recognition. ID management unit 21 to manage, input and output of stored data, memory unit 22 for storing accounts of storage that is storage, database unit 23 for DB data stored in memory unit 22, cloud Networking management unit 24 that manages the network connected to the service unit 2, the operation unit 25 that remotely performs deep learning analysis on the object for analysis, and the analysis management unit that manages the analysis result according to the operation of the operation unit 25 26 and a monitoring unit 27 for monitoring the operating state of the system, and the cloud service unit 2 may further include a security management unit 28 capable of managing security for the system.

In addition, the operation unit 25 is equipped with a virtual machine that can analyze an object through deep learning learning in the form of S/W, and a user who needs to analyze an object can access and utilize the virtual machine, characterized in that An object extraction system using cloud-based distributed deep learning information analysis and management will be provided.

delete

delete

By providing an object extraction system using the cloud-based distributed deep learning information analysis management of the present invention, large-scale facilities and costs required to process and store large amounts of data according to object analysis through deep learning learning through distributed processing and management There is a saving effect.

In addition, when a plurality of users access the network at the same time, it is possible to prevent traffic from occurring in a specific section and to more easily obtain a large amount of object information.

1 is a schematic diagram of a system according to the invention;
2 is a block diagram of the system of the present invention.
3 is an embodiment of a system according to the present invention;
4 to 6 are exemplary views according to monitoring results according to an embodiment of the present invention.
7 is an exemplary diagram illustrating various types of reports according to an embodiment of the present invention.

The object extraction system using the cloud-based distributed deep learning information analysis and management of the present invention is to solve the problem of huge data and storage when acquiring information of objects such as photos and images through deep learning learning in various fields. For this purpose, users who want to acquire various object information through distributed deep learning, centrally manage information on the storage location of objects that are analyzed and distributed after each deep learning is completed, and use object information By using only the location information of the object, it is possible to access the location where the object is stored and use the data.

Hereinafter, it will be described in detail with reference to the drawings so that those skilled in the art can easily implement the object extraction system using the cloud-based distributed deep learning information analysis management of the present invention.

1 is a schematic diagram of a system according to the present invention, FIG. 2 is a configuration diagram of a system according to the present invention, and FIG. 3 is an embodiment of a system according to the present invention.

1 to 3, the object extraction system using the cloud-based distributed deep learning information analysis management of the present invention manages cloud use application and administrator's authority management service, data backup and recovery, security, etc. In addition, the hub service unit (1) that analyzes statistics and logs and provides it in the form of a dashboard or documents it and is connected to the hub service unit (1) to create resources, create resource groups and register users, and manage security It consists of a cloud service unit (2) in charge of system infrastructure operation and management, statistics and analysis-based reporting functions.

The hub service unit 1 as described above includes a user management unit 11 that performs user subscription and authority management, an infrastructure operation management unit 12 that manages monitoring and notification functions according to data backup and storage, and system operation, and It is composed of a report management unit 13 that manages history such as backup and security log, user statistics, and security statistics.

Here, the report management unit 13 is a database unit 23 through the memory unit 22 in which the monitoring result generated by the monitoring unit 27 of the cloud service unit 2 and the analysis result of the analysis management unit 26 are stored. It is managed so that it can be provided in various types of reports using a database of

The cloud service unit 2 includes an ID management unit 21 that manages IDs for grouping, user information and recognition according to the types of objects, input and output of stored data, and a memory unit for storing accounts of storage as storage. (22), the database unit 23 that converts the data stored in the memory unit 22 into DB, the networking management unit 24 that manages the network connected to the cloud service unit 2, and the remote object for analysis It is composed of an operation unit 25 that analyzes deep learning with a arithmetic unit 25, an analysis management unit 26 that manages the analysis result according to the operation of the operation unit 25, and a monitoring unit 27 that monitors the operating state of the system.

In addition, a security management unit 28 capable of managing security for the system may be further included.

When object analysis using deep learning is required by accessing from the outside, information on object analysis is obtained by deep learning using the virtual machine of the calculation unit 25 transmitted and stored.

Here, the information may be information on the type of object, the storage location of the object for which analysis is completed, whether or not copyright is publicly known, and costs incurred due to use.

At this time, the operation unit 25 is equipped with a virtual machine capable of analyzing an object through deep learning learning in the form of S/W, and a user who needs to analyze the object can access and utilize the virtual machine.

In more detail, an object requiring analysis is analyzed through deep learning learning, and the object requiring analysis includes a video, a photo, a picture, etc., and the target object is analyzed through deep learning learning.

Here, the user can select the type of the target object for which analysis is requested. If a category according to the type of the target object is previously set, the category for each type located in the database unit 23 is selected through input of a search word on the virtual machine. It will automatically detect and notify the user of this.

In addition, when the type of the target object is an object to be analyzed for the first time, a new category can be created.

At this time, when searching for the type of the target object, the search is extended to the range of similar search terms. For example, if the target object is a mobile phone, the similar search word 'smartphone, cellular phone, cell phone, personal mobile communication terminal... ', the search is extended to the range of similar search terms, and a category is set according to the type of target object accordingly.

The analysis management unit 26 analyzes and manages the usage of H/W resources, and for this purpose, it is interlocked using Representational status transfer-API (Rest-API) so that it can be provisioned and managed programmatically.

This is to analyze the current resource usage and prevent a large amount of information from being biased on the network through appropriate distribution thereof, and to separate the server and the client and to transmit information evenly between a plurality of clients.

This is to prevent the occurrence of traffic, etc. according to the user's intensive access to some specific information.

In addition, by analyzing and managing the histories related to the user's access, information on the frequently used target object on the network is provided to a plurality of users.

In this case, a method of posting a pop-up window or a ranking may be applied as a method of providing the user with information on the target object according to the frequency of use.

The history of the analysis management unit 26 as described above is converted into a database form, stored in the memory unit 22 , and may be called and provided as information to the database 23 when necessary.

The monitoring unit 27 monitors the system, and checks the operating state of the system based on the analysis contents of the analysis management unit 26 .

In addition, if an abnormality occurs during monitoring, backup and recovery work is performed on the data of the relevant part.

At this time, the part related to the backup automatically backs up data according to a certain period set by the administrator and stores it in the memory unit 22, and when an error occurs, it is restored using the data stored in the memory unit 22. .

Monitoring is performed by the monitoring unit 27 as described above, and the results are transmitted to the report management unit 13 of the hub service unit 1 so that various types of reports can be completed.

The security management unit 28 manages security on the network, and can monitor and manage the security of user or administrator information or viruses or noise hidden in files, and network access to each cloud service. to manage the stability of

The object extraction method using the cloud-based distributed deep learning information analysis management using the system of the present invention as described above, when acquiring information of objects such as photos and images through deep learning learning in various fields, increases the size of data and In order to solve problems, various object information is acquired through distributed deep learning based on cloud services, and each deep learning is analyzed and distributed and stored information about the storage location of the object is concentrated in the center. Users who want to manage and use object information can use only the location information of the object to access the location where the object is stored and use the data.

In addition, a virtual space is formed in the database 23, the DATA-POOL for the information of the analyzed object is operated in the form of a virtual cluster, and the market is operated according to the performance of loading or using the data, and use accordingly. Depending on the performance, the user's data upload and download will be activated through the payment of points.

1: Hub service unit 2: Cloud service unit
11: User management department 12: Infrastructure operation management department
13: report management unit 21: ID management unit
22: memory unit 23: database unit
24: networking management unit 25: calculation unit
26: analysis management unit 27: monitoring unit
28: Security Management Department

Claims (6)

The object extraction system using cloud-based distributed deep learning information analysis and management manages cloud use application and administrator's permission management service, data backup and recovery, and security, and analyzes statistics and logs and provides it in dashboard form or provides documentation In order to do this, the user management unit 11, which performs user subscription and authority management, the infrastructure operation management unit 12, which manages the backup and storage of data, monitoring and notification functions according to the operation of the system, and the backup and security log, user statistics , a hub service unit (1) consisting of a report management unit (13) that manages history such as security statistics; and
In order to be connected with the hub service unit (1) and take charge of the functions of resource creation, resource group creation and user registration, system infrastructure operation and management of security management, and report functions based on statistics and analysis, according to the type of object ID management unit 21 that manages ID for grouping, user information and recognition, input and output of stored data, memory unit 22 that stores the account of storage, which is storage, stored in the memory unit 22 A database unit 23 that converts data into a DB, a networking management unit 24 that manages a network connected to the cloud service unit 2, a calculation unit 25 that remotely performs deep learning analysis on an object for analysis, a calculation unit 25 ) is composed of a cloud service unit (2) consisting of an analysis management unit (26) that manages the analysis results according to the operation of the system and a monitoring unit (27) that monitors the operating state of the system,
The calculation unit 25 is equipped with a virtual machine capable of analyzing an object through deep learning learning in the form of S/W, and when a category according to the type of the target object is established, the database ( 23), it is possible to automatically search for categories by type located in
The analysis management unit 26 is a cloud-based distributed deep learning characterized in that it uses a Rest-API (Representational status transfer-API) that can be provisioned and managed programmatically so as to analyze and manage the usage of H/W resources. Object extraction system using information analysis management.
delete The method of claim 1,
The cloud service unit (2) is an object extraction system using a cloud-based distributed deep learning information analysis management, characterized in that it may further include a security management unit (28) that can manage the security of the system.
The method of claim 1,
Cloud-based, characterized in that the operation unit 25 is equipped with a virtual machine that can analyze an object through deep learning learning in the form of S/W, and allows a user who needs to analyze an object to access and utilize the virtual machine. Object extraction system using distributed deep learning information analysis management.
delete delete

Images