KR102340644B1 - AI Type Total Security And Monitoring Service System - Google Patents

Abstract

The present invention relates to an artificial intelligence total security control service system, the configuration of which has an event (damage, scream) occurrence reporting function in response to a sound, includes a sound sensing device, and transmits a sensor measurement value message through wireless communication A wireless sound sensor that provides security functions by encrypting messages and transmits the sensor's detection status to the AI security control platform through the Internet network It performs a wired and wireless sensor interworking interface function, includes an IP gateway and SW main unit that operates for a certain period of time even after a power outage due to battery charging, performs the function of mounting and applying an AI model module, includes a device adapter, and prevents crime /AI security control platform that provides dispatch service function, access control function, disaster/safety service function, crime prevention data collection/analysis/purification function, AI learning function, AI learning result test function, AI center AI that provides AI crime prevention prediction model It includes a learning system, and the AI model further includes an AI crime prevention module and an AI disaster safety module, and the center AI learning system collects big data from the wireless sound sensor through data collection, refinement, learning, and model processes. , and the AI model module is characterized in that it expresses the dispatch probability.

Description

Artificial Intelligence Total Security Control Service System {AI Type Total Security And Monitoring Service System}

The present invention relates to an artificial intelligence total security control service system, and more specifically, the AI model module by analyzing the big data collected from the security sensor through the data collection, purification, learning, and model process of the center AI learning system. It relates to an artificial intelligence total security control service system, characterized in that it expresses the dispatch probability in

The concept of security service refers to a crime prevention and disaster prevention service that monitors and dispatches signals and information to cope with emergency situations such as crime and fire (source: Korea Information and Communication Technology Association).

1 is a conceptual diagram illustrating a physical security value chain. Referring to Figure 1, the traditional physical security control business consists of a value chain of sensors (wired sensors, controllers), networks, control systems, and control services (dispatch). By sending it, the controller generates and processes the dispatch command, and provides a service in which the dispatcher actually dispatches according to the command.

Looking at the configuration of the physical security service, as a component of the physical security service, it consists of a service and a device/terminal. Services include human (personal) security (resident security personnel, facility management and personal security), unmanned security (mechanical equipment) (sensor, CCTV detection, alarm and security personnel dispatch) and video security (CCTV, DVR, NVR, monitor, etc. monitoring) video equipment). By device/terminal type, access control (access authentication devices such as smart cards, biometrics, etc.), dispatch security (risk detection/notification devices such as alarms, alarms, sensors, etc.), and fire prevention (fire alarms, fire-fighting devices such as sprinklers) includes

The security service as described above judges whether or not to be dispatched only by the sensor on/off in the hardware (H/W) main device that accommodates the limited crime prevention sensor, so environmental impact (false alarm/error caused by wind, light, animal movement, etc.) The rate of dispatch is high. 95% of all intrusion events are false alarms).

In the existing security control service, wired sensors are used, and H/W box-type devices (installed in customers) are used as the control device (main device), the provision type is SI (Customizing) type, and simple signal monitoring and dispatch service are performed. .

On the other hand, in order to provide customized needs (crime prevention, access control, water leakage, fire, server room temperature, humidity, etc.) of corporate customers, it is a plug-and-play (ug&play) method rather than an individual system integration (SI) method. There is an urgent need for a platform that can accommodate them.

Along with the 4th industrial revolution, the Internet of Things (IoT) technology and market are showing a high growth rate. However, in the domestic market, public institutions have mainly focused on 'SmartCity' centered on CCTV integrated control and disaster safety. Recently, private conglomerates (KT/SKT/LGU+/CNS, etc.) are also developing and launching various IoT services such as HomeIoT, location tracking, and usage measurement, but an improved security service model is required.

[Prior art literature]

Republic of Korea Patent Publication No. 10-0792802 (2008.01.02) (Title of the invention: artificial intelligence crime prevention system)

An object of the present invention is to solve the problems of the prior art as described above, and when applied to physical security control based on artificial intelligence learning, reasoning, and cognitive technology, it is possible to determine whether the existing operator (controller) is an actual alarm. It is to provide a new type of artificial intelligence total security control service system that can effectively operate by preventing human error and lowering the false alarm rate by performing its role in a security system using artificial intelligence.

In addition, it is to provide an artificial intelligence total security control service system that is expected to be introduced in earnest by companies to increase productivity and increase sales thanks to the spread of machine learning and deep learning.

In addition, the center AI learning system analyzes the big data collected from the security detection sensor through data collection, refinement, learning, and model processes, and provides an artificial intelligence total security control service system that expresses the dispatch probability from the AI model module. .

An artificial intelligence total security control service system according to a preferred embodiment of the present invention for achieving the above object, in the artificial intelligence total security control service system,

A wireless sound sensor that has a function to report the occurrence of an event (damage, scream) in response to a sound, includes a sound detection device, transmits a sensor measurement value message through wireless communication, and provides a security function by encrypting the message;

Transmits the sensor's detection status to the AI security control platform through the Internet network, controls the operation of sensors and sirens/warning lights through the AI security control platform, performs wired and wireless sensor interworking interface functions, and maintains constant power even after battery charging time-acting IP gateway,

AI security control platform that includes SW main device, performs the function of loading and applying AI model module, includes device adapter, and provides crime prevention/rescue service function, access control function, disaster/safety service function,

It includes a center AI learning system that provides a crime prevention data collection/analysis/purification function, an AI learning function, an AI learning result test function, and an AI crime prevention prediction model,

The AI model is made to further include an AI crime prevention module and an AI disaster safety module,

It is characterized in that the center AI learning system analyzes the big data collected from the wireless sound sensor through data collection, refinement, learning, and model processes, and expresses the dispatch probability in the AI model module,
The center AI learning system has a common API function that implements functions commonly required for the AI security control platform, a crime prevention/response management function such as machine security and dispatch/patrol service function, and access control such as access control and attendance/ERP interworking function. There is provided an artificial intelligence total security control service system, characterized in that it includes a service API layer module that includes disaster/safety management functions such as / attendance management function, air quality, and fire detection service function.

In addition, the center AI learning system has an IoT adapter function that receives IoT packet data and delivers it to a message queue, a oneM2M adapter function that receives packet data conforming to the oneM2M standard and delivers it to a message queue, and transmits packet data from private. It is characterized by including a network layer module, characterized in that it has a non-standard adapter function that implements receiving and delivering to the message queue.

In addition, the center AI learning system receives control/collection/setting commands and provides a function to transmit control commands to the sensor through the device adapter, and the SW main device that receives the collection event from the sensor through the device adapter and implements the function to store in the DB Function, AI model function that determines whether an intrusion is detected through AI by judging intrusion information collected from sensors, and business logic functions necessary for intrusion, fire, and emergency control are implemented, sensors, access device security equipment, and customer facilities It is characterized by including a service layer module including a control/operation function that implements business logic functions necessary for operation and a collector function that receives information collected from a message queue and loads it into the DB.

delete

In addition, a crime prevention/disaster safety learning model is created using the existing machine learning algorithm or CNN (Convolutional Neural Network) algorithm or other deep learning algorithms for complex monitoring information such as heat detection, video, and access, and using the model to create a crime prevention/disaster prevention/disaster safety model It is characterized by creating a safety learning model and judging whether or not intrusion is made through the model.

In addition, the center AI learning system transmits the results obtained through the data collection, refinement, learning, and model processes to the HW main device to determine whether there is an intrusion, but the data learning model process is a process of applying the crime prevention/disaster safety learning model file. , the protocol for applying the learning model is applied, but the protocol buffer unit for applying the protocol, the weight format unit for formatting the weights in the learning model, the node unit for node setting, and the weight unit for setting the weights are operated respectively to operate the learning model It is characterized in that it creates

In addition, the center AI learning system analyzes the big data collected from the security sensor through data collection, refinement, learning, and model processes, and the AI model module is characterized by expressing the dispatch probability.

In addition, the AI crime prevention model of the AI security control platform, when an intrusion signal is detected, determines whether the AI prediction model is calculated after checking whether it belongs to the dispatch case, and basically checks the previous 9 signals when an intrusion signal occurs. (Attention layer), check the importance of the first recent signal (time t-1), and the last past signal (time t-9), multiply the weights obtained by learning, and use the CNN algorithm and GRU algorithm to extract the essential features, concatenate the feature vector extracted from the CNN algorithm and the feature vector extracted from the GRU algorithm to make one feature vector (concatenate), and calculate the final dispatch probability through the weighted sum operation on the feature vector that has been concatenated. (Dense) is characterized.

As described above, according to the artificial intelligence total security control service system according to the present invention, when applied to physical security control based on artificial intelligence learning, reasoning, and cognitive technology, the role of determining whether the existing operator (controller) is an actual alarm A new type of artificial intelligence total security control service system will be provided that enables efficient operation by preventing human error and lowering the false alarm rate by performing this in a security system using artificial intelligence.

In addition, thanks to the spread of machine learning and deep learning, it is expected that companies will introduce them in earnest to increase productivity and increase sales.

1 is an explanatory diagram illustrating a general physical security service value chain.
2 is a schematic explanatory diagram of an artificial intelligence total security control service system according to the present invention.
3 is an explanatory diagram illustrating a conventional physical security service system and a total security control service system of the present invention, respectively.
4 is a diagram illustrating the configuration of a main device and a security sensor in a conventional physical security service system.
5 is a diagram schematically showing the configuration of the main device, security sensor, and center artificial intelligence (AI) learning system in the artificial intelligence total security control service system according to the present invention.
6 is an explanatory diagram illustrating a total care service system in an artificial intelligence total security control service system according to the present invention.
7 is a schematic configuration diagram illustrating the connection relationship of each device in the artificial intelligence total security control service system according to the present invention.
8 is a diagram schematically showing the configuration of a sound detection rate measurement environment in the artificial intelligence total security control service system according to the present invention.
9 is a view showing the configuration of an intrusion event transmission accuracy measurement environment in the artificial intelligence total security control service system according to the present invention.
10 is an overall platform configuration diagram in the artificial intelligence total security control service system according to the present invention.
11 is a diagram schematically showing a machine security crime prevention service process in the artificial intelligence total security control service system according to the present invention.
12 is a diagram schematically illustrating a machine security crime prevention service process including various sensors using an IP gateway in the artificial intelligence total security control service system according to the present invention.
13 is a diagram schematically illustrating a dispatch/patrol service process in the artificial intelligence total security control service system according to the present invention.
14 is a diagram schematically illustrating an access management/access control service process in an artificial intelligence total security control service system according to the present invention.
15 is a diagram schematically illustrating a time and attendance management service process in the artificial intelligence total security control service system according to the present invention.
16 is a diagram schematically illustrating an indoor environment monitoring service process in an artificial intelligence total security control service system according to the present invention.
17 is a logical configuration diagram of the center AI learning system in the artificial intelligence total security control service system according to the present invention.
18 is a diagram schematically illustrating a data collection process in the artificial intelligence total security control service system according to the present invention.
19 is a diagram schematically illustrating a data analysis process in an artificial intelligence total security control service system according to the present invention.
20 is a diagram schematically illustrating a data purification process in an artificial intelligence total security control service system according to the present invention.
21 is a diagram schematically illustrating a learning model creation process in the artificial intelligence total security control service system according to the present invention.
22 is a diagram schematically illustrating a learning model distribution process in an artificial intelligence total security control service system according to the present invention.
23 is a diagram schematically showing an example of IP gateway operation in the artificial intelligence total security control service system according to the present invention.
24 is a diagram schematically showing an example of the configuration of an IP gateway in the artificial intelligence total security control service system according to the present invention.
25 is a diagram schematically showing an example of a wireless sound sensor service scenario in the artificial intelligence total security control service system according to the present invention.
26 is a diagram schematically illustrating a block diagram of a wireless sound sensor in an artificial intelligence total security control service system according to the present invention.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the accompanying drawings. In describing the present invention, in order to facilitate the overall understanding, the same reference numerals are used for the same components in the drawings, and duplicate descriptions of the same components are omitted.

The present invention will be described in detail with reference to examples as follows, and the scope of the present invention is not limited to the examples.

2 is a schematic explanatory diagram of an artificial intelligence total security control service system according to the present invention, FIG. 3 is an explanatory diagram illustrating a conventional physical security service system and a total security control service system of the present invention, respectively, and FIG. 4 is a conventional It is a view showing the configuration of the main device and the security sensor in the existing physical security service system, and FIG. 5 is the configuration of the main device, the security sensor, and the center artificial intelligence (AI) learning system in the total security control service system according to the present invention. 6 is an explanatory diagram illustrating a total care service system in an artificial intelligence total security control service system according to the present invention, and FIG. 7 is each in the artificial intelligence total security control service system according to the present invention. It is a schematic configuration diagram explaining the connection relationship of the devices of

8 is a view schematically showing the configuration of a sound detection rate measurement environment in the artificial intelligence total security control service system according to the present invention, and FIG. It is a view showing the configuration of the environment, and FIG. 10 is an overall platform configuration diagram in the artificial intelligence total security control service system according to the present invention, and FIG. 11 is a machine security crime prevention service process in the artificial intelligence total security control service system according to the present invention. 12 is a diagram schematically showing a machine security crime prevention service process including various sensors using an IP gateway in an artificial intelligence total security control service system according to the present invention, and FIG. 13 is a view showing the present invention It is a view schematically showing the dispatch/patrol service process in the artificial intelligence total security control service system according to the 15 is a diagram schematically showing the attendance management service process in the artificial intelligence total security control service system according to the present invention, and FIG. 16 is the indoor environment monitoring service process in the artificial intelligence total security control service system according to the present invention. is a diagram schematically showing, and FIG. 17 is a logical configuration diagram of the center AI learning system in the artificial intelligence total security control service system according to the present invention.

18 is a diagram schematically showing a data collection process in an artificial intelligence total security control service system according to the present invention, and FIG. 19 is a diagram schematically showing a data analysis process in an artificial intelligence total security control service system according to the present invention 20 is a diagram schematically showing a data purification process in the artificial intelligence total security control service system according to the present invention, and FIG. 21 schematically shows the learning model creation process in the artificial intelligence total security control service system according to the present invention 22 is a diagram schematically showing a learning model distribution process in the artificial intelligence total security control service system according to the present invention, and FIG. 23 is an IP gateway in the artificial intelligence total security control service system according to the present invention. It is a diagram schematically showing an example of operation, and FIG. 24 is a diagram schematically showing an example of the IP gateway configuration in the artificial intelligence total security control service system according to the present invention, and FIG. 25 is artificial intelligence total security control according to the present invention. It is a diagram schematically showing an example of a wireless sound sensor service scenario in a service system, and FIG. 26 is a diagram schematically showing a wireless sound sensor block in an artificial intelligence total security control service system according to the present invention.

According to the present invention, the sensor operates as a wireless sensor and a battery instead of the existing wired sensor. In addition, as a control device (main device), an intelligent S/W platform (installed on the central server) was built. In addition, the form of provision is Plug & Play, and the security control service is built as a customized TotalCareService (complex event, artificial intelligence (AI) application).

Referring to FIG. 3 , looking at the differentiation and originality of the developed technology according to the present invention compared to the existing security service, the differentiation or originality of the four types of sensor low power/wirelessization, main device SW platformization, platform intelligence, and company-customized service It has secured core technology that can provide various services by breaking through the limitations of existing physical security services.

Development of a wireless security sensor supporting low-power broadband (refer to Fig. 3)

Referring to FIG. 3 , in the conventional method, a wire for signal transmission is connected, and a wiring for power supply is required, thereby limiting the installation location ( FIG. 3 ( a ), but the method implemented through the present invention is low-power- By using Low-Power Wide-Area (LPWA) IoT technology, there is no need for wiring for power supply and transmission, so the construction cost is reduced by more than 80% compared to the basic technology (FIG. 3(b)).

On the other hand, the NB-IoT (Narrow Band IoT) technology used in the present invention is one of the LPWA Iot technologies utilizing the existing mobile communication (LTE) frequency, and is suitable for intermittent transmission of low-capacity data, such as meter reading, tracking, sensing, etc. It is mainly used and has the advantage of being able to use it for more than several years without replacing the battery due to its ultra-low power.

NB-IoT can connect up to 50,000 connections in one cell/battery life is up to 10 years or more, and the radio wave reach is 15 km.

Looking at the prospect of NB-IoT, 30 of the global operators are commercializing it in 2017, and as in the case of LTE introduction in the past, it is expected to be adopted by telecommunication companies around the world over a longer period of time.

According to the present invention, it is characterized in that the HW main device was developed as a SW main device for the first time in the world.

In other words, by configuring the main device H/W on the platform in a SW type, the efficiency of initial installation and maintenance is increased, and a customized service is provided.

The main unit (MCU H/W) is the main controller that manages all sensors in customer facilities and performs simple service processing logic.

Intelligence of security control platform applying AI algorithm

Referring to FIG. 4 , the existing method is a method in which the HW main device receives monitoring information such as heat detection, video, and access through a security sensor (security monitoring, data reception), and then performs rule-based intrusion determination, and false alarm rate This is more than 95%.

In contrast, referring to FIG. 5 , according to the present invention, by applying AI technology in a cognitive method based on empirical judgment by making the platform intelligent, the false alarm rate of intrusion is dramatically reduced by more than 50%.

In other words, it is similar in that it acquires monitoring information such as heat detection, video, and access through a crime prevention sensor, but the HW main device determines whether an intrusion is made using an artificial intelligence method (AI) rather than the existing rule-based method through crime prevention monitoring and data collection. .

That is, the sensor AI learning system transmits the results obtained through the process of data collection, refinement, learning, and model to the HW main device to apply it to determine whether there is an intrusion.

The data collection process includes crime prevention information and videos.

The data purification process includes refined learning information and images.

The data learning process is a process of creating a crime prevention/disaster safety learning model using an existing machine learning algorithm or CNN (Convolutional Neural Network) algorithm or other deep learning algorithms and learning through the model. In addition, a feed forward network (FNN), a convolutional neural network (CNN), or a recurrent neural network (RNN) may be used among deep learning algorithms.

The data learning model process is the process of applying the crime prevention/disaster safety learning model file. Here, a protocol for applying a learning model is applied, but a protocol buffer unit for applying an appropriate protocol, a weight format unit for formatting a weight in the learning model, a node unit for setting a node, and a weight unit for setting a weight are each operated. Create a learning model.

That is, in the intelligent system, a crime prevention/disaster safety learning model is created using SVM (Support Vector Machine) or CNN (Convolutional Neural Network) algorithm for complex monitoring information such as heat detection, video, and access, and intrusion through the model. By judging whether or not there is an intrusion, the false alarm rate of intrusion is reduced to about 40%.

The sensor AI learning system analyzes the big data collected from the security detection sensor through data collection, refinement, learning, and model processes to express the dispatch probability in the sensor AI learning system. The dispatch probability can be increased or decreased according to a predetermined algorithm.

Total-Care service tailored to the company

Referring to FIG. 6 , according to the present invention, a crime prevention service commonly used by corporate customers, access control system interlocking, attendance management service, and drinking water management service are basically provided.

In addition, we provide not only corporate crime prevention services, but also Total-Care services that can additionally provide customized disaster/safety, environmental monitoring, quarantine, and building/energy management for each company.

Referring to Figure 7, according to the present invention, the artificial intelligence total security control service system builds a built-in artificial intelligence (AI) security control platform on the server side, and builds a center AI learning system. In addition, on the device side, an IP gateway and a wireless sound sensor are configured.

In addition, on the PC client side, a user interface (UI) package for an administrator is provided.

In addition, in terms of the built-in AI model platform, it is composed of a SW main device, an artificial intelligence (AI) program, a crime prevention security department, an access management part, and a security sensor/IoT sensor management part.

According to the present invention, as the main performance indicator, the access recognition time is processed using a card/fingerprint reader, and as a measurement method, the user's access authentication is 1,000 ms between the access authentication device installed in the customer building and the server installed in the built-in type. Check if it is processed inside.

In addition, as a major performance indicator, intrusion detection detection time is detected by hot wire/magnet IP gateway, and intrusion event processing (heat wire + magnet + security, etc., AI model and link) is processed within 1,100ms.

In addition, as a main performance index, the sound detection rate is detected by a wireless sound sensor, and the performance index goal is to achieve a detection rate of 80% or more for window breakage and scream. As a measurement method, a wireless sound sensor and a volume meter are installed at a distance of 1m from the speaker for sound output. To check the standard volume, output the sound source and adjust the output so that the volume meter is above the standard volume. Make sure that the window break sound and scream are generated 5 times each, and that the wireless sound sensor detects it 4 times or more (evaluation item with a detection rate of 80% or more). Check whether the detected signal is received by the AI security control platform (transmission accuracy criteria evaluation item).

In addition, as a key performance indicator, the intrusion detection false alarm rate is detected by the hot wire/magnet/IP gateway, and the hot wire/access information/video information is received through the access authentication device installed in the customer building and the built-in server, and normal access Check the false alarm rate in case of an intrusion event.

In addition, as a major performance indicator, intrusion event (signal) transmission accuracy is detected through hot wire / magnet / sound sensor / IP gateway / artificial intelligence platform, and as a performance indicator target, the detection status of wired and wireless sensors connected to the IP gateway It changes and transmits it to the platform to evaluate the accuracy of whether an intrusion event occurs according to the crime prevention state (at the time of alert). As for the measurement method, we aim for a transmission accuracy of more than 99% when 100 occurrences occur in the crime prevention state (at the time of alert).

Referring to FIG. 8 , the sound detection rate measurement environment configures 5 wireless sound sensors and an IP gateway, and evaluates whether the control server detects a broken window sound or a scream. That is, after outputting the reference sound source from the demand source, the reference volume detection check, the wireless sound sensor detection check, the IP gateway reception check, and the transmission to the control server are confirmed. After that, the control server confirms the reception.

Referring to FIG. 9 , in the sound detection rate measurement environment configuration diagram, an intrusion event (signal) transmission accuracy evaluation environment is reviewed.

That is, by matching 30 wired/wireless sensors (10 heated wires / 10 magnets / 10 sound) with 5 IP G/W units, each attempts to detect intrusion more than 100 times and attempts to receive IP G/W. Evaluate whether there is a data reception error.

In addition, by matching 30 wired/wireless sensors (10 heated wires/10 magnets/10 sound) and 5 IP G/W units, each attempted intrusion detection more than 100 times, the detection signal received by IP G/W The data accuracy is evaluated by comparing the intrusion event information generated after connecting to the artificial intelligence security control platform.

Referring to Figure 10, in the artificial intelligence total security control service system according to the present invention, the built-in AI security control platform includes a SW main device, performs a function of loading and applying an AI model module, and includes a device adapter, , crime prevention/dispatch service function, access control function, and disaster/safety service function are provided.

In addition, the center AI learning system provides a crime prevention data collection/analysis/purification function, an AI learning function, an AI learning result test function, and an AI crime prevention prediction model.

The administrator UI performs dashboard, access/security management function, history statistics inquiry function, environment setting function, and operation management function.

The intrusion detection AI model includes an intrusion detection security control model and a safety/disaster security control model.

On the other hand, the IP gateway transmits the detection status of the sensor to the platform through the Internet network, controls the operation of the sensor and siren/light light by the platform, performs various wired and wireless sensor interworking interface functions, and is constant even after a power outage by charging the battery. time will work.

The wireless sound sensor has a function to report event occurrence in response to sound (breakage, scream, etc.) It transmits the sensor measurement value as a message and provides a security function by encrypting the message.

Referring to FIG. 10 , as a built-in AI security control platform, an architecture will be schematically described.

The network layer module has an IoT adapter function that receives IoT packet data and delivers it to a message queue, a oneM2M adapter function that receives packet data conforming to the oneM2M standard and delivers it to a message queue, and receives packet data from a private and sends a message It has a non-standard adapter function that implements delivery to the queue. Parts not described in the drawings are only an adapter option of the device adapter module and will not be described separately.

The service layer module receives control/collection/setting commands and implements the function to transmit control commands to the sensor through the device adapter, receives the collection event from the sensor through the device adapter, and implements the storage function in the DB, and the sensor AI model function that implements the function to determine the intrusion through AI by determining intrusion information collected from the It includes a control/operation function that implements the business logic functions required for equipment and customer facility operation, and a collector function that implements the function to receive information collected from the message queue and load it into the DB.

The service API Layer module includes a common API function that implements functions commonly required for the platform, a crime prevention/dispatch management function that implements machine security and dispatch/patrol service functions, and an access function that implements access control and attendance/ERP interworking functions. /Includes disaster/safety management function that implements /time and attendance management function, air quality, and fire detection service function.

11 is a view showing an example of a service through the total security control service system according to the present invention, and the machine security crime prevention service is intrusion by judging whether or not intrusion is made in the artificial intelligence platform when a signal is generated after a detector (magnet, heating wire, etc.) is installed. The signal is received from the integrated control center (local), and a dispatch command is issued to the local person in charge (dispatched).

12 is a machine security crime prevention service including various sensors using an IP gateway, and a detector (wired/wireless magnet, wired/wireless heated wire, wired/wireless emergency bell, infrared, shock, shutter, glass, safe, fire, etc.) After installation, when an intrusion occurs, the detection zone and detection status are transmitted to the platform, the AI platform judges whether there is an intrusion, displays the danger with a warning light linked to the IP gateway, and receives the intrusion signal from the integrated control center (local), the local person in charge (Dispatch) gives an order to dispatch.

13 is a dispatch/patrol service, after installing a detector (magnet, heating wire, etc.), when a signal is generated, the AI platform determines whether an intrusion has occurred, and the intrusion signal is simultaneously received from the integrated control center (local) and the central control. Dispatched personnel are dispatched from the 1st local integrated control setter, and in the event of a failure in the 1st local area, the 2nd integrated control center (central) is dispatched to the site of the dispatching personnel as a remote location.

14 is an access management/access control service, which opens the door only to a user authorized with an access control reader to block crimes from outsiders in advance.

15 is a time and attendance management service that manages employee time and attendance through local control, opens the door only to users authorized with an access control reader, and allows access control in a specific area through 52-hour service management. Execute door opening and operation management within time.

Figure 16 relates to an indoor environment monitoring service, receiving detection of a dangerous (screaming sound / glass breaking sound) sound source in a specific area through a wireless sound sensor, and determining the emergency risk through artificial intelligence learning according to standards such as fluctuation range / rate of change So, dispatch personnel to the site to deal with it.

17 is a schematic diagram of the center AI learning system. The AI model layer module 32 consists of an AI crime prevention model and an AI disaster / safety model, and the function of applying the AI model of the AI security control platform through the distribution component carry out

The Service Layer module 34 performs a function of collecting, analyzing, refining, and learning security monitoring information in connection with the Service API 26 of the AI security control system 20 .

The Core Layer module 36 performs a function of processing or storing information for learning in common.

The resource management layer module 38 performs a function of managing the Data Layer distributed file resource.

The Data Layer module 39 is composed of collected information storing information such as access, image, sound detection, heat detection, and location of the AI security control platform and learning data for learning.

The AI security control platform 20 includes a network layer module 22 , a service layer module 24 , and a service API layer module 26 . The service layer module 24 includes a SW main unit, an AI crime prevention model, and an AI disaster safety module. The service API layer module 26 receives the security/disaster safety AI model distributed in the AI model layer 32 of the center AI learning system, and receives access, image, gas concentration, heat detection, location, equipment maintenance rate, It transmits the presence or absence of dispatch and equipment attribute values to the collection unit of the service layer 34 of the AI learning system.

Although not shown, the AI security model of the AI security control platform 20 may perform the following functions, for example. In other words, when an intrusion signal is received at the control center (Alert) (that is, the signal is detected by the security/monitoring sensor 10), after checking whether it belongs to the dispatch case defined by the controller know-how, it is checked whether the AI prediction model is calculated. It determines (Rule Based Layer) (the AI crime prevention model decides), and basically considers the context by checking the previous 9 signals together when an intrusion signal occurs (Attention layer) (the AI crime prevention model considers).

However, the importance of the first recent signal (time t-1) and the last past signal (time t-9) among the 9 signals is different. Relatively, the signal just before the intrusion signal (time t-1) should be observed more carefully than the last past signal (time t-9). Therefore, the attention layer multiplies the previous 9 signals by weights according to their importance, and the corresponding weights are obtained through learning.

Among the previous signals, there are features that must be captured regardless of the point of view. For example, when a service facility is repairing equipment, it is important to quickly find the AS entry signal among the previous 9 signals because various signals, including intrusion signals, are transmitted to the control center through security/monitoring sensors during the inspection process (CNN Block ). The CNN block plays a role in extracting such essential features. The panel data that has passed through the attention layer is considered context and features are extracted (GRU Block). For example, when an intrusion signal occurs during the morning rush hour, there is a high probability that it is a customer error in the case of a facility that habitually occurs during the same time period. However, in the case of a facility that has never made a mistake before, the probability of dispatch should be high.

After that, the process of concatenating the feature vector extracted from the CNN block and the feature vector extracted from the GRU block into one feature vector (concatenate) is performed.

Calculate the probability of the last occurrence through the weighted sum operation on the feature vector that has been concatenated (Dense).

On the other hand, access, image, gas concentration, heat detection, etc. are made from the security/monitoring sensor 10 .

On the other hand, the functions in the Core Layer module 36 and the resource management layer module 38 are not important in the present invention and are optional, so that they are not described in more detail. In the present invention, in the data layer module 39, which is a set of collected information and learning data, in connection with the Service API 26 of the AI security control system 20, the function of collecting, analyzing, refining, and learning security monitoring information is performed. The Service Layer module 34 performs a service function.

Referring to FIG. 18 , in the AI learning system function, the data collection function collects security monitoring information (access, image, sound detection, heat detection, location, dispatch, etc.) using the Service API of the AI security control platform. and performs the function of storing collected information through the Spark component. However, this is an exemplary process.

Referring to FIG. 19 , in the AI learning system function, the data analysis function analyzes security monitoring information in various ways using the R Tool of Spark Core to create an AI crime prevention/disaster safety learning data set (Pearson, Kendall, Spear) Only correlation coefficient analysis, etc.) can be performed. However, this is only an example, and in the present invention, a predictive model was created and used using the TensorFlow library.

Referring to FIG. 20 , in the AI learning system function, the data purification function refines the data (missing values, noise, errors) so that the AI learning module (Spark MLIB, Tensorflow, etc.) can learn without errors through the results of data analysis. information processing). This is an exemplary process.

Referring to FIG. 21 , in the AI learning system function, the learning model generation function generates an AI crime prevention model and an AI disaster safety model using a crime prevention learning data set and a disaster safety learning data set created through data purification.

Referring to Figure 22, in the AI learning system function, the learning model distribution function uses the SW main device's OpenAPI to transfer the learning model (crime prevention / disaster learning model) created to enable intelligent crime prevention and disaster services to the SW main device. It performs the function of distributing to

In consideration of the demand for new sensors other than the sensors operated in the present invention, a device for transmitting the detection values of various wired/wireless sensors to the AI platform is required, and a sound sensor device for selectively reacting and detecting a designated sound source is required.

IP gateway development

IP Gateway Overview

When an event sensor is detected (heat ray detection, door open, impact, etc.), an IP gateway device is a device that transmits a signal to the platform and notifies an alarm (warning light/siren) according to the designation of the platform in order to understand the situation before and after sensor detection.

Service scenario using IP gateway

Referring to FIG. 23 , an example of IP gateway operation can be seen. As shown in FIG. 23, ① when the heat sensor detects a human body, an event is transmitted to the IP gateway, ② the IP gateway transmits an intrusion signal among the detection signals to the platform, ③ the platform checks the intrusion signal and transmits it to the control; ④ Transmits a warning light/siren alarm to the IP gateway, and ⑤ the dispatch command is issued from the control.

Looking at the IP gateway function and main characteristics, it transmits sensor detection information to the platform through the Internet, is in charge of device operation control (siren, etc.) It is easy to add a sensor, and it operates for a certain period of time even in a power outage by charging the battery.

The IP gateway application field is indoor environments such as offices, buildings, factories, and workshops, and it can be applied to the security field and dangerous areas that require quick response to disasters, and can be applied to places where it is necessary to build a sensor network composed of various sensors.

IP Gateway (GW:Gateway) product configuration

Referring to FIG. 23, an IP gateway block diagram is shown. The MCU operates with the optimal function and current consumption for IoT, performs PHY module control, message generation, transmission/reception, and analysis functions, sensor detection, measurement, It performs control functions, performs external power, battery status detection and report functions, console input/output processing, status LED display control, auxiliary sensor function, case open detection, and battery voltage detection functions.

The PHY Module is a TCP/IP function module and has an Ethernet interface provision function.

Power Drive and Battery perform battery charging function when using external power source (phase power), uninterrupted service provision function when external power is cut off and recovery, and provide notification in case of low voltage by measuring battery status.

As an input/output device, it is possible to check the operation status by the LED display in the LED display unit, and has a function of providing a terminal monitor and a console for S/W update.

wireless sound sensor

It is used for a service that reports event detection in response to a specific sound such as breakage or scream, and transmits the detection result in a wireless communication method so that you can respond to an event/accident according to the control of the platform.

Wireless Sound Sensor Service Scenario

As shown in FIG. 25 , ① the wireless sound sensor transmits a specific sound detection result to the platform, and ② the platform determines the detection result and issues an alarm. ③ If necessary, instruct evacuation.

Looking at the functions and main features of the wireless sound sensor, it reports event occurrence in response to a specific sound, attaches it to the measurement point without separate wiring, and has excellent cost-saving effect due to convenient installation/relocation, and maximum battery usage due to low power function Time can be secured, periodic reporting by setting and reporting cycle change according to specific fluctuation range/rate is possible It provides a security function by encryption.

In the field of wireless sound sensor application, offices, warehouses, stores, and homes are places that require security, and it can be applied to areas where accidents and accidents occur frequently, and it can also be applied to women's safe shoulder roads and public toilets.

The detailed specification of the wireless sound sensor is as follows.

- 8bit u-controller unit, IoT purpose processor

- Active analytic sound detector

- Low Power Consumption Regulator

- Support Power Save Mode

- event-driven light weight software

- encryption AES256 of messages from/to platform

Referring to FIG. 26 , as a block diagram of a wireless sound sensor, the MCU operates with an IoT-specific optimal function and current consumption, has sensor detection and measurement functions, and provides sound sensor control, message generation, transmission, message reception, and interpretation functions. It has a battery status detection and reporting function, a console input/output processing function, and a status LED display control function.

Sound Detector has a specific sound input function, a specific sound detection function, and a power control function.

The Power Drive and Battery can provide the maximum amount of power per unit volume, and an auxiliary circuit that can provide the maximum amount of current is employed, and the battery status is measured to provide a notification when the voltage is low.

As an input/output device, the LED provides an operation status and detection status check function by LED display, a connector for sensor selection is provided, and a terminal monitor and management console are provided.

As described above, the best embodiment has been disclosed in the drawings and the specification. Although specific terms are used herein, they are used only for the purpose of describing the present invention, and are not used to limit the meaning or scope of the present invention described in the claims. Therefore, it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible therefrom. Accordingly, the true technical protection scope of the present invention should be defined by the technical spirit of the appended claims.

10: security / monitoring sensor
20: AI Security Control Platform
30: Center AI Learning System

Claims (7)

In the artificial intelligence total security control service system,
A wireless sound sensor that has a function to report the occurrence of an event (damage, scream) in response to a sound, includes a sound detection device, transmits a sensor measurement value message through wireless communication, and provides a security function by encrypting the message;
Transmits the sensor's detection status to the AI security control platform through the Internet network, controls the operation of sensors and sirens/warning lights through the AI security control platform, performs wired and wireless sensor interworking interface functions, and maintains constant power even after battery charging time-acting IP gateway,
AI security control platform that includes SW main device, performs the function of loading and applying AI model module, includes device adapter, and provides crime prevention/rescue service function, access control function, disaster/safety service function,
It includes a center AI learning system that provides a crime prevention data collection/analysis/purification function, an AI learning function, an AI learning result test function, and an AI crime prevention prediction model,
The AI model is made to further include an AI crime prevention module and an AI disaster safety module,
The center AI learning system analyzes the big data collected from the wireless sound sensor through data collection, refinement, learning, and model processes, and the AI model module expresses the dispatch probability,
The center AI learning system has a common API function that implements functions commonly required for the AI security control platform, a crime prevention/response management function such as machine security and dispatch/patrol service function, and access control such as access control and attendance/ERP interworking function. Artificial intelligence total security control service system, characterized in that it includes a service API layer module that includes disaster/safety management functions such as / attendance management function, air quality, and fire detection service function. The method of claim 1, wherein the center AI learning system has an IoT adapter function that receives IoT packet data and delivers it to a message queue, a oneM2M adapter function that receives packet data conforming to the oneM2M standard and delivers it to a message queue, Private Artificial intelligence total security control service system, characterized in that it includes a network layer module, characterized in that it has a non-standard adapter function that receives packet data from and delivers it to a message queue. The method of claim 1, wherein the center AI learning system receives control/collection/setting commands and provides a function to transmit a control command to the sensor through a device adapter, and receives a collection event from the sensor through a device adapter and stores it in the DB It implements the SW main device function to implement, the AI model function that judges the intrusion information collected from the sensor and determines whether there is an intrusion through AI, and implements the business logic function necessary for intrusion, fire, and emergency control, and secures the sensor and access device Characterized in that it includes a service layer module including a control/operation function that implements business logic functions necessary for equipment and customer facility operation, and a collector function that receives information collected from a message queue and loads it into the DB Artificial intelligence total security control service system. delete The method according to claim 1, wherein a crime prevention/disaster safety learning model is generated using an existing machine learning algorithm, a convolutional neural network (CNN) algorithm, or other deep learning algorithms for complex monitoring information such as heat detection, image, and access, and the model is used. An artificial intelligence total security control service system, characterized in that it creates a crime prevention/disaster safety learning model and judges whether or not intrusion occurs through the model. According to claim 1, wherein the center AI learning system transmits the results obtained through the data collection, purification, learning, and model process to the HW main device to determine whether there is an intrusion, but the data learning model process uses the crime prevention/disaster safety learning model file As an application process, a protocol for applying a learning model is applied, but a protocol buffer unit for applying the protocol, a weight format unit for formatting a weight in the learning model, a node unit for node setting, and a weight unit for setting a weight are each An artificial intelligence total security control service system, characterized in that it operates and creates a learning model. According to claim 1, wherein the AI security model of the AI security control platform, when an intrusion signal is detected, after checking whether it belongs to the dispatch case, determines whether to calculate the AI prediction model, and basically, when an intrusion signal occurs, the previous nine signals Considering the context by checking together (Attention layer), checking the importance of the first recent signal (time t-1), and the last past signal (time t-9), multiplying the weights obtained by learning, CNN algorithm and The essential features are extracted using the GRU algorithm, and the feature vector extracted from the CNN algorithm and the feature vector extracted from the GRU algorithm are concatenated to make one feature vector (concatenate), and the final dispatch is performed through the weighted sum operation on the feature vector that has been concatenated. Artificial intelligence total security control service system, characterized in that calculating the probability (Dense).

Images