Disaster reporting system

Abstract

To provide a disaster reporting system that promptly reports to a pre-registered recipient when an emergency occurs at a pre-registered facility.SOLUTION: A disaster reporting system 100 includes: a posted information determination system 110 that determines, based on content data which is at least one of image data and a message text indicating the content of posted information posted on the Internet 400, contained within the posted information, whether or not the posted information indicates a fire or similar incident at a registered facility where the posted information is pre-registered; and a reporting unit 122 that, if the posted information determination system 110 determines that the posted information indicates a fire or similar incident at a registered facility, reports to a report destination terminal pre-registered as a report destination for fire occurrences at the registered facility that a fire has occurred.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a disaster notification system that quickly notifies a designated notification destination of the occurrence of a fire or the like in a pre-registered facility. [Background technology]

[0002] In recent years, with the development of social networking services (SNS), various events are posted on the Internet along with images and message text.

[0003] Here, the following system has been disclosed as a system for grouping information posted on SNS (see, for example, Patent Document 1). The image information display system according to Patent Document 1 uses machine learning to identify the type of event, location, and date and time of the event shown in posted images related to any incident automatically collected from the Internet. The image information display system then groups the posted images by incident based on the identified items, and generates a display screen. The image information display system then displays the generated display screen on a display device. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-164788 Summary of the Invention [Problem to be solved by the invention]

[0005] For companies that operate facilities throughout Japan and overseas, if a fire breaks out at one of their local facilities, the facility may be so busy extinguishing the fire that the initial report to the head office may be delayed. In this case, the head office may only learn of the fire at their facility through television news, which may delay their response.

[0006] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a disaster reporting system that quickly reports to a pre-registered reporting destination when an abnormal situation occurs at a pre-registered facility. [Means for solving the problem]

[0007] The disaster reporting system of the present disclosure includes a determination unit that determines whether posted information posted on the Internet indicates an abnormal situation at a pre-registered registered facility based on content data, which is at least one of image data and message text indicating the content of the posted information, and a reporting unit that, if the determination unit determines that an abnormal situation is occurring at the registered facility, reports the occurrence of an abnormal situation to a reporting destination terminal that is pre-registered as a reporting destination in the event of an abnormal situation occurring at the registered facility. [Effects of the Invention]

[0008] According to the present disclosure, it is possible to obtain a disaster notification system that quickly notifies a pre-registered notification destination when an abnormal situation occurs at a pre-registered facility. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of an overall overview of a system according to a first embodiment of the present disclosure. [Figure 2] 2 is a block diagram showing an example of the configuration of the disaster notification system shown in FIG. 1. FIG. [Figure 3] FIG. 2 is a diagram showing management information held by the disaster notification system. [Figure 4] FIG. 10 is a flowchart illustrating the operation of the disaster notification system. DETAILED DESCRIPTION OF THE INVENTION

[0010] A preferred embodiment of the disaster notification system of the present disclosure will be described below with reference to the drawings. In the following description, a fire will be described as an example of an abnormal situation, but the present disclosure can also be applied to natural disasters such as earthquakes, floods, tornadoes, and typhoons. The disaster notification system of the present disclosure is a system that acquires posted information about the occurrence of a fire via a social networking service (SNS).

[0011] In addition, the disaster notification system disclosed herein has a technical feature that if the acquired posted information includes posted information indicating that a fire has occurred at a pre-registered facility, it will promptly notify the contact point associated with that facility.

[0012] Embodiment 1 FIG. 1 is a diagram illustrating an example of an overall overview of a system according to a first embodiment of the present disclosure.

[0013] In FIG. 1, a disaster notification system 100 is connected to an SNS system 200 and a destination terminal 300 via the Internet 400 .

[0014] The SNS system 200 is a system that manages information posted on the Internet 400 .

[0015] The SNS system 200 is a system that forms an SNS network that accepts posted information from pre-registered members and publishes this posted information on the Internet 400. Note that the SNS system 200 may also be a community site such as a mailing list or an electronic bulletin board. The SNS system 200 may also be a site that collects and publishes posted information of a specific category, for example, information related to disasters, from among the posted information posted on the Internet 400.

[0016] When information is posted on the Internet 400 that a disaster such as a fire has occurred at a pre-registered facility, the disaster notification system 100 notifies the notification destination associated with the facility.

[0017] The disaster notification system 100 includes a posted information determination system 110 and a customer management system 120. In the first embodiment, the posted information determination system 110 and the customer management system 120 are configured as separate systems, but they can also be implemented as a single system.

[0018] The posted information determination system 110 functions as a determination unit, and determines whether or not the posted information indicates a fire at a registered facility that has been registered in advance, based on image data included in the posted information.

[0019] The customer management system 120 is a system that manages the registration information of contracted companies that use the disaster notification system 100. As will be described later, the customer management system 120 registers information about the facilities of the contracted companies and information about the contact points designated by the contracted companies.

[0020] Furthermore, if a fire breaks out at the facility, the customer management system 120 notifies the contact point designated by the contracted company.

[0021] The destination terminal 300 is a terminal to which a report is to be sent in the event of a fire. The destination terminal 300 is a smartphone, a personal computer, or the like provided by each contracted company as a destination for a report.

[0022] FIG. 2 is a block diagram showing an example of the configuration of the disaster notification system 100 shown in FIG.

[0023] 2, the posted information determination system 110 includes a registered facility classification unit 111 and a fire site classification unit 112.

[0024] The registered facility classification unit 111 acquires posted information posted on the Internet 400. Then, the registered facility classification unit 111 classifies the acquired posted information to determine whether it is posted information about a registered facility.

[0025] The fire site classification unit 112 acquires posted information that has been classified as a registered facility by the registered facility classification unit 111 from the registered facility classification unit 111. Then, the fire site classification unit 112 classifies the posted information acquired from the registered facility classification unit 111 to determine whether it is posted information about the site of a fire.

[0026] Based on the classification results of the registered facility classification unit 111 and the fire site classification unit 112, the posted information determination system 110 can determine whether the posted information indicates a fire at a registered facility that has been registered in advance.

[0027] 2, the customer management system 120 includes a registration unit 121 and a reporting unit 122.

[0028] The registration unit 121 registers and centrally manages information provided by contracted companies. The registration unit 121 also receives requests from contracted companies to update the registered information and performs update processing in accordance with the instructions. The information managed by the registration unit 121 will be described later.

[0029] When the posted information determination system 110 determines that a fire has broken out at a registered facility, the reporting unit 122 reports the information to the reporting destination associated with the registered facility. This allows the reporting destination terminal 300 within the contracted company to quickly receive a report that a fire has broken out at a facility related to the company based on the posted information.

[0030] Furthermore, the destination terminal 300 can check the posted information about the fire based on the report, and can visually understand the scale and state of the fire from the image data.

[0031] FIG. 3 is a diagram showing management information managed by the registration unit 121. As shown in FIG.

[0032] Management information 140 is a table that associates registered facilities, which are facilities of each company, with the destination to report to in the event of a fire at each registered facility. Management information 140 stores, as attribute information for each registered facility, a facility ID for uniquely identifying the registered facility, the company name to which the registered facility belongs, the facility name, facility address, and facility coordinate values ​​of the registered facility. Management information 140 also stores, as information on the destination to report, a destination telephone number and a destination email address.

[0033] In addition to this management information 140, image data of each facility to be detected is also registered in the registration unit 121. For example, the image data may be image data of each facility photographed from a public road or the like at the eye level of an ordinary passerby. In addition, multiple image data of each facility photographed from various angles are registered.

[0034] Fig. 4 is a diagram illustrating a flowchart showing the operation of the disaster notification system 100. The process shown in Fig. 4 is repeatedly executed.

[0035] In step S101, the registered facility classification unit 111 acquires posted information that has been made public on the Internet 400. Specifically, the registered facility classification unit 111 acquires posted information from the SNS system 200 using an API (Application Programming Interface) provided by the operating company that operates the SNS system 200.

[0036] It should be noted that, with a large amount of various posted information being made public every day, randomly acquiring this information and classifying it as information related to a fire site in the subsequent fire site classification unit 112 would require an excessive amount of calculation. For this reason, the registered facility classification unit 111 may acquire posted information by narrowing it down to categories related to fires, as follows:

[0037] For example, the registered facility classification unit 111 may acquire mainly posted information with a hash tag related to fires, or mainly acquired posted information posted by senders who mainly send disaster information. Furthermore, when acquiring posted information, the registered facility classification unit 111 may add a time condition, such as limiting the acquired posted information to data posted within one hour from the current time.

[0038] The acquisition conditions that can be set depend on the API provided by the operating company of the SNS system 200 .

[0039] In step S102, the registered facility classification unit 111 acquires image data from the posted information.

[0040] To acquire the image data, the registered facility classification unit 111 first acquires the posted information ID assigned to the posted information from the SNS system 200 using the API function.

[0041] Then, the registered facility classification unit 111 uses this posted information ID to obtain detailed information about the posted information from the SNS system 200. This detailed information includes a URL for accessing the image data, and the registered facility classification unit 111 obtains the image data by referencing this URL.

[0042] The message text posted by the poster is included as text data in the acquired detailed information. The registered facility classification unit 111 can also use this message text in classification processing, as will be described later.

[0043] In this way, the registered facility classification unit 111 acquires the image data contained in the posted information. Note that this method is merely an example, and the method for acquiring the image data depends on the configuration of the SNS system 200, the API provided, etc.

[0044] In step S103, the registered facility classification unit 111 performs a registered facility classification process. The registered facility classification process is a process for classifying posted information to determine whether it is posted information about a registered facility, as shown below.

[0045] When processing image data, the registered facility classification unit 111 performs image analysis based on a registered facility learning model that has been trained using registered facilities as detection targets.

[0046] Here, the registered facility learning model is a model trained using image data of each registered facility taken under normal circumstances as training data. Furthermore, during training, the training data is taken from the expected line of sight of the poster, i.e., the direction and position from which a passerby would view the facility. This allows for the creation of a highly accurate registered facility learning model.

[0047] In addition, the registered facility classification unit 111 can compare the posted image data with the image data of each facility registered in the registration unit 121 to calculate the similarity, and based on the similarity, classify the posted information as whether it is posted information about a registered facility.

[0048] In this case, the registered facility classification unit 111 may acquire image data of each facility from the registration unit 121 each time it performs the registered facility classification process, or may store image data of each facility in advance inside the registered facility classification unit 111. Alternatively, feature amounts for comparison may be extracted for each piece of image data of a facility registered in the registration unit 121, and these feature amounts may be stored in advance inside the registered facility classification unit 111 instead of the image data.

[0049] If the registered facility classification unit 111 classifies the posted information as posted information about a registered facility in step S104, the process proceeds to step S105. On the other hand, if the registered facility classification unit 111 classifies the posted information as not posted information about a registered facility, the process in FIG. 4 ends.

[0050] In step S105, the registered facility classification unit 111 outputs the facility ID of the classified facility. The registered facility classification unit 111 holds in advance the facility ID of each facility to be classified, and if the posted information is a registered facility, it can output the facility ID of the classified facility as the classification result.

[0051] In step S106, the fire scene classification unit 112 performs a fire scene classification process.

[0052] When processing image data as part of the fire scene classification process, the fire scene classification unit 112 classifies the posted information as to whether it is information about a fire scene based on a fire scene learning model that has been trained using fire scenes as detection targets. In step S105, the fire scene classification unit 112 performs classification processing on posted information that has been classified as information about a registered facility.

[0053] The fire scene learning model is a model that is trained using image data of images of fire scenes as training data. When building the fire scene learning model, the model builder obtains image data of various fire scenes from the Internet 400 and uses this as training data. Alternatively, the model builder may obtain a set of image data of fire scenes collected by a third party and use this as training data. Alternatively, the model builder may obtain a pre-built learning model specialized for detecting fires from a third party and use this as the fire scene learning model.

[0054] If the fire site classification unit 112 classifies the posted information as posted information about a fire site in step S107, the process proceeds to step S108. On the other hand, if the fire site classification unit 112 classifies the posted information as not posted information about a fire site, the process in FIG. 4 ends.

[0055] In step S108, the posted information determination system 110 transmits the facility ID as a processing result to the reporting unit 122 of the customer management system 120. The facility ID transmitted here is the facility ID output in step S105.

[0056] In step S109, the reporting unit 122 of the customer management system 120 acquires the facility ID transmitted from the posted information determination system 110. Then, the reporting unit 122 searches the management information 140 using the acquired facility ID as a search key, and acquires the facility name, facility address, report destination telephone number, and report destination email address.

[0057] The notification unit 122 notifies the destination telephone number by SMS (Short Message Service) and also notifies the destination email address by email. At this time, the notification unit 122 notifies that a fire has occurred at the facility, including the facility name and address in the body of the email.

[0058] In addition, when multiple posters who have confirmed the same fire each post individually, the reporting unit 122 will end up making multiple reports about the same fire. To avoid this, the reporting unit 122 may refrain from reporting the same facility for a while after making a report once. Alternatively, the reporting unit 122 may make a report when individual posts about the same fire are made a predetermined number of times within a predetermined time period.

[0059] In the explanation of the process in Figure 4, the steps of collecting posted information, classifying registered facilities, and classifying fire sites are shown as one flow, but these can also be processed in parallel, and a process of organizing related information posted to SNS in chronological order can also be performed in parallel, so that the notification recipient can refer to the organized information. Note that with regard to the notification step, there is no need to notify again of a fire that has already been notified, but as the amount of organized information increases over time, the display of the organized information at the notification recipient can be updated.

[0060] As described above, Fig. 4 has been explained with fires as the target, but it is also possible to target abnormal situations such as events that occur artificially or naturally at a facility, disasters, etc. Natural disasters such as earthquakes, floods, tornadoes, and typhoons can also be treated as abnormal situations, but it is preferable to target events that affect individual facilities rather than disasters that have a wide impact on the region.

[0061] <Other implementation examples> In the first embodiment, the posted information determination system 110 classifies image data into whether it is a registered facility or a fire scene. Meanwhile, the posted information includes not only the image data but also message text posted by the poster as content data indicating the content of the posted information.

[0062] Here, we will explain an implementation example in which various classifications are performed based on the message text posted by the poster.

[0063] When classifying whether a message is a registered facility by processing the message text, the registered facility classification unit 111 checks whether the message text contains the character strings of the company name, facility name, and address shown in Fig. 3. This allows the registered facility classification unit 111 to classify whether the posted information is a registered facility. Here, characters contained in the image data as text may be identified and treated as message text.

[0064] In this case, the character strings of the company name, facility name, and address shown in FIG.

[0065] Alternatively, the registered facility classification unit 111 may perform language analysis using a registered facility learning model trained using the character strings of the company names, facility names, and addresses shown in Fig. 3 as training data, thereby enabling the registered facility classification unit 111 to classify posted information as being a registered facility.

[0066] Furthermore, when classifying whether or not a message is a fire site by processing the message text, the fire site classification unit 112 checks whether or not specific characters such as "fire," "fire," "flame," "burn," etc. are included in the message text. This allows the fire site classification unit 112 to classify whether or not the posted information is information indicating a fire.

[0067] Furthermore, when processing message text, the fire scene classification unit 112 can also classify, by machine learning, whether the posted information is about the scene of a fire. The learning model used at this time is constructed by training all past posted messages posted when a fire broke out using as training data.

[0068] When processing a message text, the fire scene classification unit 112 performs language analysis by using this trained model as a fire scene training model.

[0069] This allows the fire scene classification unit 112 to classify the posted information as being about the scene of a fire, based on the message text.

[0070] Furthermore, when the posted information includes coordinate information indicating the location where the information was posted or the location where the image was taken, the registered facility classification unit 111 may classify whether the posted information is about a registered facility based on the coordinate information. In this case, the registered facility classification unit 111 can classify whether the posted information is about a registered facility by comparing the facility coordinate values ​​shown in FIG. 3 with the coordinate information included in the posted information.

[0071] The posted information determination system 110 may also perform a process to eliminate fake information. This elimination of fake information can be achieved by using a specified logic to determine whether the image data of the posted information has been manipulated, or by accumulating past image information as training data and comparing it to determine whether the information is fake.

[0072] The above-described aspects may be combined with each other. That is, the disaster notification system 100 may process content data, which is at least one of image data and message text indicating the content of posted information. In addition, although the above description assumes that the registered facility classification unit 111 performs processing before the fire site classification unit 112 performs processing, this order may be reversed.

[0073] The features of such a disaster notification system 100 can be summarized as follows: the system has the following configuration and can achieve the following effects.

[0074] The disaster notification system 100 includes a posted information determination system 110 that determines whether posted information posted on the Internet 400 indicates an abnormal situation at a registered facility that has been registered in advance, based on content data that is at least one of image data and message text that indicates the content of the posted information.

[0075] In addition, the disaster notification system 100 is equipped with a notification unit 122 that, when the posted information determination system 110 determines that an abnormal situation has occurred at a registered facility, notifies the notification destination terminal 300 that has been pre-registered as the destination to report to in the event of an abnormal situation occurring at the registered facility that an abnormal situation has occurred.

[0076] Therefore, the disaster notification system 100 can utilize information posted via SNS as a means of detecting abnormal situations. Therefore, when an abnormal situation occurs in a facility, the disaster notification system 100 can quickly report the situation to pre-registered notification destinations.

[0077] Furthermore, when the content data is image data, the posted information determination system 110 classifies the posted information as being about a registered facility based on a registered facility learning model that has been trained using registered facilities as detection targets. Then, based on the results of this classification, the posted information determination system 110 determines whether the posted information indicates an abnormal situation at the registered facility. In this case, the registered facility learning model is a model that has been trained using image data captured of the registered facility under normal circumstances as training data.

[0078] Therefore, the classification accuracy of registered facilities can be improved compared to when training data is not used.

[0079] Furthermore, when the content data is image data, the posted information determination system 110 classifies the posted information as whether it is posted information about the site where an abnormal event occurred, based on an abnormal event site learning model that has been trained using the site where an abnormal event occurred as a detection target. Then, the posted information determination system 110 determines whether the posted information indicates an abnormal event at a registered facility based on the results of the classification. In this case, the abnormal event site learning model is a model that has been trained using image data of the site where an abnormal event occurred as training data.

[0080] Therefore, the accuracy of classification of the scene of an abnormal event can be improved compared to when training data is not used.

[0081] The posted information includes coordinate information, and the posted information determination system 110 determines whether the posted information indicates the occurrence of an abnormal situation at a registered facility based on the coordinate information.

[0082] This makes it possible to improve the accuracy of the determination.

[0083] In this way, the business operator of the disaster notification system according to the first embodiment can develop an information provision service that determines abnormal situations at pre-registered facilities, etc. based on information posted on the Internet, and promptly notifies pre-registered contact points of the occurrence of the abnormal situation.

[0084] On the other hand, users of the disaster notification system according to the first embodiment can centrally manage information about abnormal situations such as fires occurring at pre-registered facilities in a wide area, including overseas, as a service that provides rapid information via a route separate from the automatic fire alarm system using fire detectors, fire receivers, etc. installed at each facility.

[0085] For example, consider a case where an automobile manufacturer is a user of the disaster notification system according to the first embodiment, and a disaster prevention-related equipment manufacturer is a business operator of the disaster notification system according to the first embodiment.

[0086] In this case, users register in advance not only their own factories, branch offices, and other facilities, but also the facilities of companies that make up the supply chain, such as parts suppliers and companies to which they outsource subcontract work. This allows users to receive information more quickly as an information provision service based on information posted on the Internet before it is announced on television or in other media, even if the company's information gathering system is delayed when an abnormal situation occurs at one of the facilities and causes disruptions to production.

[0087] On the other hand, disaster notification system operators can cooperate with their customers to provide a service that provides information about abnormal situations such as fires at pre-registered customer facilities, including facilities that do not have their own automatic fire alarm equipment, all in one place based on information posted on the Internet, thereby supporting crisis management for their customers' businesses. [Explanation of symbols]

[0088] 100 Disaster reporting system, 110 Posted information determination system (determination unit), 111 Registered facility classification unit, 112 Fire scene classification unit, 120 Customer management system, 121 Registration unit, 122 Reporting unit, 140 Management information, 200 SNS system, 300 Report destination terminal, 400 Internet.

Claims

1. a determination unit that determines whether or not posted information posted on the Internet indicates an abnormal situation at a registered facility that has been registered in advance, based on content data that is at least one of image data and message text that indicate the content of the posted information, which is included in the posted information; a reporting unit that, when the determination unit determines that an abnormality has occurred in the registered facility, reports the occurrence of the abnormality to a reporting destination terminal that is pre-registered as a reporting destination when an abnormality has occurred in the registered facility; A disaster notification system equipped with:

2. When the content data is image data, The determination unit classifying the posted information as to whether it is posted information about the registered facility based on a registered facility learning model that has been trained using the registered facility as a detection target; Based on the result of the classification, it is determined whether the posted information indicates an abnormal situation at the registered facility; The registered facility learning model is a model trained using image data of the registered facility taken under normal circumstances as training data. The disaster notification system according to claim 1 .

3. When the content data is image data, The determination unit classifying the posted information as to whether it is posted information about the site where an abnormal event has occurred based on an abnormal event site learning model that has been learned using the site where an abnormal event has occurred as a detection target; Based on the result of the classification, it is determined whether the posted information indicates an abnormal situation at the registered facility; The abnormal situation scene learning model is a model trained using image data captured at the scene of an abnormal situation as training data. The disaster notification system according to claim 1 .

4. The posted information includes coordinate information, The determination unit determines whether the posted information indicates the occurrence of an abnormal situation at the registered facility based on the coordinate information. The disaster notification system according to claim 1 .

5. The determination unit is configured by a posted information determination system, and the posted information determination system includes a registered facility classification unit and a fire site classification unit, the registered facility classification unit acquires the posted information after narrowing it down to a category related to a fire, and the fire site classification unit classifies the posted information to determine whether it is posted information about a fire site. The disaster notification system according to claim 1 .

Images