Information processing system, and method for controlling the information processing system

Abstract

To reduce the costs associated with using AI services. [Solution] An information processing system according to one aspect of the present disclosure includes an information acquisition unit 203 that acquires input information D1 relating to a monitored object, a request unit 204 that instructs a generation AI service to perform processing by transmitting instruction information D2 relating to the input information D1 acquired by the information acquisition unit 203 to the generation AI service, and a request frequency reduction processing unit 206 that performs a first determination process to determine whether processing is necessary based on the success or failure of a first condition that increases the need for processing. When the request frequency reduction processing unit 206 determines that processing by the generation AI service is necessary, the request unit 204 requests processing from the generation AI service.

Description

Technical Field

[0001] The present disclosure relates to an information processing system and a control method for an information processing system.

Background Art

[0002] Patent Document 1 discloses a technique for determining the presence or absence of damage caused by a disaster based on an image captured by an imaging device when the disaster is detected. Patent Document 2 also discloses a technique for determining a disaster state from a disaster video using a deep learning unit that has learned the disaster state.

[0003] In recent years, AI (Artificial Intelligence) services that enable the use of AI such as the deep learning unit in Patent Document 2 through a network have become known.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] Generally, a fee is charged for using an AI service provided by a vendor. When the fee system is a volume-based charging system, a charge is incurred for the amount of use of the AI service. Therefore, for example, when configuring a system that constantly causes an AI service to determine a disaster state in order to monitor the occurrence of a disaster in Patent Document 2, there is a problem that the charge amount increases according to the length of the monitoring period and the determination frequency of the disaster state during the monitoring period, and the cost associated with using the AI service increases. This problem is a common problem for any system that uses an AI service. This disclosure aims to provide an information processing system that can reduce the costs associated with using AI services, and a method for controlling such an information processing system. [Means for solving the problem]

[0006] An information processing system according to one aspect of this disclosure includes: an information acquisition unit that acquires input information relating to a target to be monitored; a request unit that instructs a generation AI service to perform processing by transmitting instruction information relating to the input information acquired by the information acquisition unit; and a request frequency reduction processing unit that performs a first determination process to determine whether the processing is necessary based on the success or failure of a first condition that increases the need for the processing, wherein the request frequency reduction processing unit determines that the generation AI service is necessary to perform the processing, and the request unit requests the generation AI service to perform the processing.

[0007] A control method for an information processing system according to one aspect of the present disclosure includes: a first step of acquiring input information relating to a target to be monitored; a third step of instructing a generation AI service to perform processing by transmitting instruction information relating to the input information acquired in the first step; and a second step of executing a first determination process to determine whether the processing is necessary based on the success or failure of a first condition that increases the need for the processing, wherein if the second step determines that the processing by the generation AI service is necessary, the third step requests the generation AI service to perform the processing. [Effects of the Invention]

[0008] According to one aspect of this disclosure, the costs associated with using AI services can be reduced. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows the configuration of a monitoring system according to an embodiment of this disclosure. [Figure 2] This diagram shows the functional configuration of a cloud information processing server. [Figure 3]This flowchart shows the operation of the cloud information processing server. [Figure 4] This is a flowchart for the process of reducing the frequency of requests. [Figure 5] This is a flowchart of the request frequency reduction process related to the first modified example of this disclosure. [Figure 6] This figure shows the functional configuration of the cloud information processing server relating to the second modified example of this disclosure. [Figure 7] This figure shows an example of input information. [Figure 8] This is a diagram illustrating the operation of the removal processing unit. [Figure 9] This flowchart shows the operation of the cloud information processing server according to the third modified example of this disclosure. [Figure 10] This diagram shows the configuration of the monitoring system relating to the fourth modified example of this disclosure. [Figure 11] This is a flowchart showing the operation of the cloud information processing server according to the fourth modified example of this disclosure. [Modes for carrying out the invention]

[0010] The following describes preferred forms relating to this disclosure with reference to the drawings. Note that the dimensions and scale of parts in the drawings may differ from actual dimensions as appropriate, and some parts may be shown schematically for ease of understanding. Furthermore, unless otherwise specified in the following description to limit the scope of this disclosure, the scope of this disclosure is not limited to the forms described below. The scope of this disclosure includes equivalents of such forms.

[0011] Figure 1 shows the configuration of the monitoring system 1 according to this embodiment. The monitoring system 1 is a system that monitors one or more predetermined states in one or more objects (hereinafter referred to as "monitored objects"). The monitoring system 1 of this embodiment is one form of an information processing system that performs various processes related to the objects using AI services available via a communication network.

[0012] In this embodiment, a cloud AI service is used as the AI service. A cloud AI service is a service that makes AI resources built on a public communication network NW such as the Internet available to users as needed and in the necessary amounts. For example, Amazon Web Services (AWS), Microsoft Azure, and Google Cloud are well-known. Generally, a cloud AI service is provided by a cloud server realized by distributed computing of a plurality of physical servers. Note that the public communication network NW is an aspect of a communication network.

[0013] In this embodiment, the cloud AI service is provided有偿 by a vendor, and a pay-per-use system is adopted for the pricing system. The pay-per-use system is a pricing system in which a charge corresponding to the usage amount of the cloud AI service is incurred.

[0014] As shown in FIG. 1, the monitoring system 1 includes a cloud generation AI server 10, a cloud information processing server 20, and one or more (a plurality in the illustrated example) user-side systems 30.

[0015] The cloud generation AI server 10 is a cloud server that provides a cloud generation AI service and includes a generation AI service unit 101. The generation AI service unit 101 is a functional unit that provides a cloud generation AI service via a public communication network NW or the like. The cloud generation AI service is an aspect of both a cloud AI service and a mere AI service. The generation AI used in the cloud generation AI service is also referred to as generative AI and is an AI capable of generating various contents such as text, images, and voices. For the learning model of such generative AI, a model that has previously learned the characteristics and relationships of each of a vast amount of data (big data) by deep learning is used.

[0016] It should be noted that the word "有偿" in the original text seems incorrect. I translated it as "有偿" first according to the text, but it may be a misspelling. It might be "有偿" which means "for a fee" or "charged". If this is a mistake in the original, please correct it for a more accurate translation.In this embodiment, when the generation AI service unit 101 receives input information D1 about a monitored object and instruction information D2 that instructs the determination of a predetermined state in the monitored object, it determines the predetermined state of the monitored object based on the input information D1 in accordance with the instructions of the instruction information D2, and outputs information related to the predetermined state (hereinafter referred to as "output information D3"). Input information D1 corresponds to the first information in this disclosure. Output information D3 is, for example, information relating to the determination result of the predetermined state. In order to provide such a generation AI service, the generation AI service unit 101 is equipped with a learning model that has learned a vast amount of data about the monitored object using deep learning.

[0017] Furthermore, in this embodiment, the AI ​​of the generation AI service unit 101 is a so-called multimodal AI that performs judgment by combining input information D1 of different information modalities (formats), such as text, images, audio, video, and detection signals from various sensors.

[0018] The user-side system 30 is a user-side system that utilizes the monitoring system 1, and as shown in Figure 1, it comprises an information gathering device 301, a communication device 302, and a user device 303.

[0019] The information gathering device 301 is a device that continuously or intermittently at predetermined intervals collects input information D1 about the monitored object. In this embodiment, examples of monitored objects include roads, inside buildings, and outdoors. Examples of predetermined states in the monitored object include the occurrence of an emergency, a critical situation, or an abnormal situation. Hereinafter, emergency, critical situation, or abnormal situation will be collectively referred to as a "specific situation." Examples of specific situations on roads include the occurrence of flooding or traffic accidents. Examples of specific situations inside buildings include the occurrence of a fire or a worker collapsing unconscious. Examples of specific situations outdoors include the occurrence of a disaster. Another example of a specific situation in an outdoor location, such as a park, is the detection of ball games. This specific situation is used to understand and warn about ball damage to surrounding houses caused by ball games in parks.

[0020] The information gathering device 301 includes a camera that, for example, captures still or moving images of roads, inside buildings, or outdoors, in order to collect current information about the monitored object as input information D1.

[0021] The communication device 302 comprises a transmitting device 302A and a receiving device 302B. The transmitting device 302A is a device that transmits the input information D1 collected by the information collection device 301 to the cloud information processing server 20 via the public communication network NW. The transmission frequency of the input information D1 is arbitrary. That is, the transmission frequency may be the same as the frequency at which the information collection device 301 collects the input information D1, or it may be less frequent. The receiving device 302B is a device that receives monitoring information D4 from the cloud information processing server 20 via the public communication network NW. The monitoring information D4 includes information relating to a predetermined state at the monitored target, and includes, for example, output information D3 from the cloud generation AI server 10 and information such as alerts that notify of the occurrence of a specific situation at the monitored target. The communication device 302 may also be provided in the user device 303. Furthermore, monitoring information D4, including alerts and other information, sent from the cloud information processing server 20, may be directly received by an appropriate terminal of the administrator responsible for monitoring the monitored target. In addition, the monitoring system 1 may be equipped with a notification server that sends notifications such as email, and the cloud information processing server 20 may send monitoring information D4 to an appropriate terminal of the administrator via the notification server.

[0022] The user device 303 includes an output device 303A that outputs monitoring information D4 in a manner recognizable by the user. The output device 303A may include, for example, a display that shows text or image information, a speaker that outputs sound, and a lamp whose illumination state changes according to the monitoring information D4. In addition to the monitoring information D4, the output device 303A may also output the input information D1 collected by the information collection device 301. Furthermore, the user device 303 may be equipped with an operating device that accepts user operations. For example, a personal computer may be used as the user device 303.

[0023] The cloud information processing server 20 is a server managed by the vendor providing the monitoring system 1, and is a cloud server that implements communication functions equivalent to the functions of the communication device of the physical server, storage functions equivalent to the functions of the storage device of the physical server, and arithmetic processing functions equivalent to the functions of the arithmetic processing unit of the physical server. Furthermore, the cloud information processing server 20 is one embodiment of the information processing device in this disclosure.

[0024] Figure 2 shows the functional configuration of the cloud information processing server 20. As shown in the figure, the cloud information processing server 20 includes a communication unit 201, a user management unit 202, an information acquisition unit 203, a request unit 204, a monitoring information generation unit 205, and a request frequency reduction processing unit 206.

[0025] The communication unit 201 is a functional unit that communicates with the user-side system 30 via the public communication network NW, and also communicates with the cloud-generated AI server 10. The functions of the communication unit 201 are realized by the above-mentioned communication functions.

[0026] The user management unit 202 is a functional unit that pre-stores user management data 202A. User management data 202A is data that records the monitoring target and predetermined status for each user. The functions of the user management unit 202 are realized by the above-mentioned storage function.

[0027] The information acquisition unit 203 is a functional unit that continuously or intermittently acquires input information D1 related to the monitored object, which is received by the communication unit 201. The functions of the information acquisition unit 203 are realized by the above-mentioned arithmetic processing function. Furthermore, the information acquisition unit 203 is a functional unit that corresponds to the first information acquisition unit in this disclosure.

[0028] The request unit 204 is a functional unit that requests the cloud-generated AI server 10 to make a judgment about a predetermined state in the monitored object. The function of the request unit 204 is realized by the calculation processing function described above. In this embodiment, the request unit 204 requests the cloud generation AI server 10 to make a decision by inputting the input information D1 of the monitored target and the instruction information D2 that instructs the cloud generation AI server 10 to make a decision regarding a predetermined state in the monitored target. Text information is used for the instruction in the instruction information D2. The text information is textual information called a prompt that gives instructions or requests to the generation AI service. In this embodiment, text information is created in advance to cause the cloud generation AI server 10 to make an appropriate decision based on the input information D1 regarding a predetermined state in the monitored target (i.e., the state in which a specific situation occurs), depending on the combination of the monitored target, the predetermined state, and the information modality (format) and content of the input information D1. This text information is recorded in the instruction statement data 204A.

[0029] Since a generative AI service is used as the AI ​​service, a common generative AI service can be used even when making judgments about different predetermined states (for example, flooded conditions and traffic accident conditions). In addition, by simply changing the instruction information D2 as appropriate, it is possible to determine different predetermined states from the same monitored input information D1 (e.g., captured images).

[0030] Furthermore, because the AI ​​used in the generation AI service employs multimodal AI, compared to so-called single-modal AI that performs decision-making based on a predetermined type of input information D1, for example, it can perform flexible and advanced decision-making processes by appropriately changing the text information of instruction information D2 or by using multiple input information D1 from different information modals.

[0031] In this embodiment, in order to eliminate ambiguity from the determination result of the occurrence of a specific situation which is a predetermined state, the text information of instruction information D2 uses a question format that asks for a response with two options: "YES" or "NO".

[0032] For example, if the object being monitored is a road, the specific event is the occurrence of flooding, and the input information D1 is a photograph of the road, the text information in instruction information D2 would be the sentence, "Is the road shown in the photograph flooded?" For example, if the object being monitored is a road, the specific event is the occurrence of a traffic accident, and the input information D1 is a photograph of the road, then the text information in instruction information D2 would be the sentence, "Has a traffic accident occurred on the road shown in the photograph?" For example, if the object being monitored is inside a building, the specific incident is that a worker has fainted and collapsed, and input information D1 is a photograph taken inside the building, then the text information in instruction information D2 would be the sentence, "Is there a person lying down in the photograph?" For example, if the object being monitored is a parking lot, the specific situation is the vacancy status of the parking lot, and input information D1 is a photograph of the parking lot, then the text information in instruction information D2 would be the phrase "Are there any vacant parking spaces?".

[0033] These question-format text information are used in instruction information D2, and output information D3 containing either "YES" or "NO" answers to the question of whether a specific situation has occurred is obtained from the cloud-generated AI server 10.

[0034] In Figure 2, the monitoring information generation unit 205 is a functional unit that generates the monitoring information D4 based on the output information D3 output by the cloud-generated AI server 10. The monitoring information generation unit 205 also has the function of controlling the transmission of the monitoring information D4 from the communication unit 201 to the user-side system 30. The functions of the monitoring information generation unit 205 are realized by the calculation processing function described above.

[0035] The request frequency reduction processing unit 206 is a functional unit that performs request frequency reduction processing to reduce the frequency of requests made by the request unit 204. The request frequency reduction processing will be described later. The function of the request frequency reduction processing unit 206 is realized by the calculation processing function described above.

[0036] The cloud information processing server 20 may be built on the same cloud as the cloud generation AI server 10. Furthermore, the cloud information processing server 20 may be a physical server instead of a cloud server. Furthermore, each of the aforementioned functional units provided by the cloud information processing server 20 is realized by having the cloud server execute a predetermined program. This predetermined program can be recorded on a recording medium and distributed, or distributed via telecommunications.

[0037] Figure 3 is a flowchart showing the operation of the cloud information processing server 20. When the information acquisition unit 203 receives input information D1 from the user-side system 30 (step Sa1), the request frequency reduction processing unit 206 executes the request frequency reduction process described above (step Sa2). The request frequency reduction process determines whether or not a request for decision processing to be made to the cloud generation AI server 10, and a more detailed explanation will be given later.

[0038] If the processing in step Sa2 determines that a request is not necessary (i.e., unnecessary) (step Sa3: NO), the processing procedure returns to step Sa1, and the system waits for the next input information D1. In this way, if the request frequency reduction processing unit 206 determines that a request is unnecessary, no request is made to the cloud generation AI server 10, thus reducing the amount charged under the pay-per-use system, i.e., the cost of using the cloud generation AI server 10.

[0039] If the processing in step Sa2 determines that a request is necessary (step Sa3: YES), the request unit 204 requests the cloud generation AI server 10 to perform a decision (step Sa4). In detail, the processing in step Sa4 is as follows: First, the request unit 204 identifies the monitoring target and predetermined state associated with the user of the user-side system 30 that sent the input information D1, based on the user management data 202A (step Sa4-1). Through the processing in step Sa4-1, even if the monitoring system 1 monitors different monitoring targets for each user, the appropriate monitoring target and predetermined state are identified for each user who sent the input information D1.

[0040] Next, the request unit 204 refers to the instruction data 204A and identifies the text information to be used in the instruction information D2 based on the monitoring target associated with the user, the predetermined state, and the information modality of the input information D1, and generates the instruction information D2 containing the said text information (step Sa4-2). Then, the request unit 204 sends the input information D1 and the instruction information D2 to the cloud generation AI server 10 to request a determination process for the predetermined state of the monitoring target (step Sa4-3).

[0041] When step Sa4 is performed, the cloud generation AI server 10 determines a predetermined state in the monitored object based on the input information D1, in accordance with the instructions in the text information of instruction information D2, and outputs the output information D3, which is information related to that predetermined state. In this embodiment, as described above, the predetermined state is the state in which a specific situation has occurred, and the output information D3 will include information in which the answer to the question of whether or not a specific situation has occurred is "YES" or "NO". This output information D3 is transmitted from the cloud generation AI server 10 to the cloud information processing server 20.

[0042] When the communication unit 201 of the cloud information processing server 20 receives output information D3, the monitoring information generation unit 205 generates monitoring information D4 based on the output information D3 and transmits the monitoring information D4 from the communication unit 201 to the user-side system 30 (step Sa5). In step Sa5, if the output information D3 contains a "NO" response indicating that the specific incident has not occurred, the monitoring information generation unit 205 includes information indicating that the specific incident has not occurred in monitoring information D4. On the other hand, if the output information D3 contains a "YES" response indicating that the specific incident has occurred, the monitoring information generation unit 205 includes alert information in monitoring information D4 to notify the user of the occurrence of the specific incident.

[0043] As a result of the processing in step Sa5, monitoring information D4 is output from the user device 303 of the user-side system 30. Furthermore, if a specific situation occurs in the monitored system, an alert is issued from the user device 303 based on the alert information contained in monitoring information D4. This alert allows the user to quickly understand that a specific situation has occurred in the monitored system and to take appropriate countermeasures.

[0044] Steps Sa4 and Sa5 are repeatedly executed at predetermined intervals to continuously monitor for the occurrence of specific events. This interval is set appropriately for each specific event at the monitored location. For example, if the monitored location is a road and the specific event is flooding, the interval is set to once every hour. For example, if the monitored location is a road and the specific event is a traffic accident, the interval is set to once every minute. For example, if the monitored location is inside a building and the specific event is a worker losing consciousness, the interval is set to once every minute.

[0045] Figure 4 is a flowchart of the request frequency reduction process described above. As shown in Figure 4, the request frequency reduction processing unit 206 first determines whether the first condition is met (step Sb1: first determination process). The first condition is a condition that increases the need for decision processing by the cloud-generated AI server 10. For example, the first condition is a condition that increases the likelihood of a predetermined state in the monitored object changing. In this embodiment, the first condition is a condition that increases the likelihood of a specific event occurring in the monitored object.

[0046] For example, if the object of monitoring is a road and the specific event is the occurrence of flooding, considering that flooding is more likely to occur during a certain period when the amount of rainfall in a day is relatively large, the first condition can be "the present time is within the period of that specified time (for example, between April and October)."

[0047] For example, if the object being monitored is a road and the specific event is the occurrence of a traffic accident, considering that unusual noises usually occur when a traffic accident occurs, a microphone can be installed in the user-side system 30, and the first condition can be "the occurrence of an unusual noise."

[0048] For example, if the monitoring target is a building where toxic gases may be generated, and a specific situation is when a worker inside the building faints, then, considering that the worker's fainting is caused by the generation of toxic gases, the first condition could include "toxic gases being detected by a gas detector," "workers being detected inside the building by a motion sensor," and "fire alarms being activated."

[0049] For example, if the object of monitoring is inside a building and the specific event is the occurrence of a fire, then the first condition can be "the fire alarm has activated." For example, if the object of monitoring is outdoors and the specific condition is the occurrence of a disaster, the first condition can be "an emergency alert is issued by a public agency in connection with the occurrence of a disaster."

[0050] Furthermore, the information required to determine whether the first condition is met or not can be obtained from appropriate information sources included in the user-side system 30, or from appropriate information sources other than the user-side system 30. That is, the cloud information processing server 20 has a function to acquire information from information sources via the communication unit 201 or other appropriate functional unit, and the request frequency reduction processing unit 206 determines whether the first condition is met or not based on such information. Information sources include, for example, various sensors installed in the user-side system 30, and various information sources that transmit information via the public communication network NW.

[0051] Next, as shown in Figure 4, if the first condition is not met (step Sb1: NO), the request frequency reduction processing unit 206 determines that no decision processing by the cloud generation AI server 10 is necessary (step Sb2). On the other hand, if the first condition is met (step Sb1: YES), the request frequency reduction processing unit 206 determines that decision processing by the cloud generation AI server 10 is necessary (step Sb3).

[0052] This request frequency reduction process ensures that only when there is a high probability of a specific situation occurring within the monitored area, it is determined that a decision-making process by the cloud-generated AI server 10 is necessary, and a request for decision-making processing is sent to the cloud-generated AI server 10. In other words, the fulfillment of the first condition triggers a request for decision-making processing to the cloud-generated AI server 10. Therefore, unnecessary requests to the cloud-generated AI server 10 are prevented, and the costs associated with using the cloud-generated AI server 10 are also reduced.

[0053] As described above, the monitoring system 1 of this embodiment includes a cloud information processing server 20. The cloud information processing server 20 includes an information acquisition unit 203 that acquires input information D1 related to the monitored object, a request unit 204 that instructs processing by transmitting instruction information D2 related to the input information D1 acquired by the information acquisition unit 203 to a cloud generation AI server 10 that provides generation AI services, and a request frequency reduction processing unit 206 that executes a first determination processing to determine whether the processing is necessary based on the success or failure of a first condition that increases the need for the processing. When the request frequency reduction processing unit 206 determines that processing by the generation AI service is necessary, the request unit 204 requests processing from the cloud generation AI server 10.

[0054] Therefore, requests are only made to the cloud-based AI server 10 providing the AI ​​generation service when there is a high need for processing using the AI ​​generation service. This effectively reduces unnecessary requests and lowers the costs associated with using the AI ​​generation service.

[0055] Furthermore, according to this embodiment, since a generation AI service is used, a common generation AI service can be used even when determining the occurrence of different specific events (for example, the occurrence of flooding and the occurrence of a traffic accident). In addition, by simply changing the instruction information D2 as appropriate, it is possible to determine the occurrence of different specific events from the same monitored input information D1 (for example, captured images).

[0056] Furthermore, according to this embodiment, multimodal AI is used in the AI ​​of the generation AI service. Therefore, compared to a so-called single-modal AI that performs decision-making based on a predetermined type of input information D1, for example, it can perform flexible and advanced decision-making by appropriately changing the text information of instruction information D2 or by using multiple input information D1 from different information modals.

[0057] The embodiments illustrated above can be modified in various ways. Specific examples of modifications that can be applied to the aforementioned embodiments are given below. Two or more embodiments arbitrarily selected from the following examples can be combined as appropriate, to the extent that they do not contradict each other.

[0058] (First variation) Figure 5 is a flowchart of the request frequency reduction process related to the first modified example. In this figure, the same reference numerals are used for the same steps as in Figure 4, and the explanation of those steps is omitted. As shown in Figure 5, in the request frequency reduction process of the first modified example, if the first condition is met (step Sb1: YES), the request frequency reduction processing unit 206 determines whether the second condition is met (step Sc1: second determination process). The request frequency reduction processing unit 206 then determines that if the second condition is met (step Sc1: YES), a determination process by the cloud generation AI server 10 is necessary (step Sb3), and if the second condition is not met (step Sc1: NO), a determination process by the cloud generation AI server 10 is unnecessary (step Sb2).

[0059] The second condition is a condition in which the need for decision-making processing by the cloud-generated AI server 10 becomes higher than when the first condition is met. In this modified example, the second condition is a condition in which the possibility of a specific situation occurring in the monitored area becomes even higher. For example, if the object being monitored is a road and the specific event is flooding, the first condition might be "the period is one in which the amount of rainfall in a day is relatively high," and the second condition might be "the weather forecast for that day is rain" or "it is currently raining."

[0060] Furthermore, the cloud information processing server 20 is equipped with a function to acquire information required for determining the success or failure of the second condition, in the same way as the information required for determining the success or failure of the first condition, from the user-side system 30 or an appropriate information source other than the user-side system 30, via the communication unit 201 or other appropriate functional unit.

[0061] According to the first modified version, a request is made to the cloud-generated AI server 10 only when the need for processing using the cloud-generated AI server 10 has increased significantly. Therefore, the costs associated with using the cloud-generated AI server 10 can be further reduced.

[0062] In the first modified example, the request frequency reduction processing unit 206 may determine whether three or more conditions are met that indicate a gradually increasing need for decision processing by the cloud-generated AI server 10, and if all conditions are met, it may determine that decision processing by the cloud-generated AI server 10 is necessary.

[0063] Furthermore, in the first modified example, if it is possible to determine whether the second condition is met or not from the input information D1, the request unit 204 may request the cloud generation AI server 10 to make such determination (i.e., the processing in step Sc1 in Figure 5). In this case, it is preferable that the frequency of requesting the determination of the second condition is less than or equal to the frequency of requesting the determination of the occurrence of a specific event in the monitored object.

[0064] (Second variation) Figure 6 shows the functional configuration of the cloud information processing server 20 according to the second modified example. In this figure, the same reference numerals are used for the same functional blocks as in Figure 2, and the explanation of those functional blocks is omitted.

[0065] As shown in Figure 6, the cloud information processing server 20 according to the second modified example differs in configuration from the cloud information processing server 20 according to the embodiment in that it includes an information processing unit 207. The information processing unit 207 is a functional unit that, before a request is made to the cloud generation AI server 10 by the request unit 204, applies appropriate processing to the input information D1 to improve the accuracy of the decision processing using the generation AI service when sending the input information D1 as is might lead to a decrease in the accuracy of the decision processing using the generation AI service. The information processing unit 207 is implemented, for example, by the above-mentioned calculation processing function. In this modified example, as an example of processing, a method is described in which the information processing unit 207 removes information from the input information D1 that would cause a decrease in the accuracy of the generation AI service.

[0066] Figure 7 shows an example of input information D1. In the example shown in the figure, input information D1 is information for the case where the monitored object is a road and the specific event is flooding on the road, and includes information including a photograph taken of the monitored road. When the monitored object is outdoors, it is often illuminated by lighting equipment during the night after sunset. In this case, as shown in Figure 7, although the illuminated area A that is lit by the lighting equipment is brightly captured in the photograph, the area around the illuminated area A is dark due to insufficient light. Therefore, when a photograph taken at night is input as input information D1 to the cloud generation AI server 10 and the cloud generation AI server 10 is asked to determine the state of flooding on the road, the illuminated area A may be mistakenly recognized as the flooded area. Due to this misrecognition, the cloud generation AI server 10 mistakenly judges that flooding has occurred even when it has not actually occurred, and as a result, an alarm is output from the user device 303 of the user-side system 30.

[0067] To avoid this situation, the information processing unit 207 removes images outside the illumination range A from the captured image. Specifically, as shown in Figure 8, the information processing unit 207 extracts the effective range B set within the illumination range A by cropping, thereby generating a captured image as new input information D1 from which the light-insufficient areas around the illumination range A have been removed. This prevents misrecognition and misjudgment by the cloud-generated AI server 10, and prevents a decrease in the accuracy of the judgment process.

[0068] Furthermore, if the input information D1 may cause a decrease in the accuracy of the judgment processing using the generated AI service only under specific circumstances, the information processing unit 207 processes the input information D1 only while those specific circumstances are occurring. In the example in Figure 7, whether or not the specific circumstances are occurring can be determined from the time of day, the lighting status of the road lighting fixtures, etc. Furthermore, the processing performed by the information processing unit 207 may not only remove information that would reduce the accuracy of the judgment processing using the generation AI service, but may also involve adjusting the brightness, contrast, and color tone of the image.

[0069] (Modified version of 2A) In the embodiment described above, the request unit 204 may add an instruction to the instruction information D2 indicating that it will perform processing based on the information processed by the information processing unit 207 from the input information D1 described in the second modified example. This prevents misrecognition and misjudgment by the cloud-generated AI server 10, similar to the second modification, and prevents a decrease in the accuracy of the judgment process.

[0070] (Third variation) Figure 9 is a flowchart showing the operation of the cloud information processing server 20 according to the third modified example. In this figure, the same reference numerals are used for the same steps as in Figure 3, and the explanation of those steps is omitted. As shown in Figure 9, the request unit 204 of the cloud information processing server 20 in the third modified example requests the cloud generation AI server 10 to perform a decision (step Sa4), and then further requests the cloud generation AI server 10 to perform other processing related to the decision result regarding the occurrence of a specific situation (step Sd1). For example, if the request unit 204 requests a decision in step Sa4 regarding "the occurrence of flooding on the road," then in step Sd1 it requests a decision on "whether there are people or cars around the flooded area, and whether those people or cars are submerged." According to this operation, if the cloud generation AI server 10 determines, through the processing in step Sa4, that flooding has occurred in the monitored area, the cloud generation AI server 10 will determine whether or not the situation is such that actual damage could occur due to the flooding. Based on these determinations, the user can gain a deeper understanding not only of the occurrence of a specific incident in the monitored area, but also of the various events that may arise as a result of that incident.

[0071] (Fourth variation) Figure 10 shows the configuration of the monitoring system 1 according to the fourth modified example of this disclosure. In this figure, components identical to those in Figure 1 are denoted by the same reference numerals, and their descriptions are omitted. The monitoring system 1 according to the fourth modified example, as shown in Figure 10, includes two types of cloud-generated AI servers: a cloud-generated AI server 10 and a low-cost cloud-generated AI server 40. The low-cost cloud-generated AI server 40, like the cloud-generated AI server 10, includes a generation AI service unit 401 and provides cloud-generated AI services. In this embodiment, the processing accuracy of the generation AI used in the cloud-generated AI service of the low-cost cloud-generated AI server 40 is lower than that of the generation AI of the cloud-generated AI server 10. Generally, differences in the processing accuracy of generation AI arise from differences in the amount of training data, the number of parameters in the training model, and the processing power of the cloud server itself. In this embodiment, because the processing accuracy of the low-cost cloud-generated AI server 40 is lower, the cost associated with using the cloud-generated AI service is set lower than that of the cloud-generated AI server 10.

[0072] Figure 11 is a flowchart showing the operation of the cloud information processing server 20 according to the fourth modified example. In this figure, the same reference numerals are used for the same steps as in Figure 3, and the explanation of those steps is omitted. As shown in Figure 11, when the cloud information processing server 20 determines that a request is necessary to the cloud generation AI server 10 (step Sa3: YES), the request frequency reduction processing unit 206 instructs the low-cost cloud generation AI server 40 to make the decision, rather than the cloud generation AI server 10, to request the request from the request unit 204 (step Se1). Steps Se1-1 and Se1-2 included in this step Se1 are the same as steps Sa4-1 and 4-2 shown in Figure 3. Step Se1-3 is the same as step Sa4-3 except that the destination of the input information D1 and instruction information D2 (i.e., the request destination) is the low-cost cloud generation AI server 40.

[0073] Next, the request frequency reduction processing unit 206 determines, based on the output information D3 output from the low-cost cloud generation AI server 40, whether or not the low-cost cloud generation AI server 40 has determined that a specific situation has occurred in the monitored area (step Se2). If it is determined that no specific incident has occurred (Step Se2: NO), the processing procedure returns to Step Se1-2, and requests are repeatedly made to the low-cost cloud-generating AI server 40 in order to continue monitoring for the occurrence of a specific incident in the monitored area. On the other hand, if it is determined that a specific situation has occurred (Step Se2: YES), the request frequency reduction processing unit 206 requests the cloud-generated AI server 10, which has higher judgment accuracy, to perform the judgment processing in order to ensure the accuracy of the judgment (Step Sa4-3).

[0074] According to the fourth modification, a low-cost cloud-generated AI server 40 is primarily used for decision-making processes using the cloud-generated AI service, thus further reducing the costs associated with using the cloud-generated AI service. In addition, if the low-cost cloud-generated AI server 40 determines that a specific situation has occurred, the same decision is requested from the cloud-generated AI server 10, which performs more accurate decisions, thus preventing misjudgments regarding the occurrence of the specific situation.

[0075] (Fifth variation) In the embodiment described above, the case in which the input information D1 includes a captured image was illustrated. However, the input information D1 may include multiple pieces of information of different information modalities (for example, image information and audio information). This allows the cloud-generated AI server 10 to perform more advanced processing and improves processing accuracy.

[0076] (Sixth variation) In the embodiment described above, if the image, which is input information D1, contains a memory for measuring a physical quantity related to a predetermined state, the instruction information D2 may instruct the system to determine the predetermined state based on the value of that memory. For example, if the object being monitored is a road and the specific event is flooding, a memory device can be installed on a wall or utility pole facing the road so that it can be photographed using a camera, and the water level of the road can be measured from the memory device.

[0077] (Seventh variation) In the embodiment described above, the cloud-generated AI server 10 was made to determine whether or not a specific situation had occurred. However, the cloud-generated AI server 10 may also be made to determine whether or not such a specific situation is likely to occur, that is, to make a future prediction. [Explanation of Symbols]

[0078] 1...Monitoring system, 10...Cloud generation AI server, 20...Cloud information processing server, 30...User-side system, 40...Low-cost cloud generation AI server, 203...Information acquisition unit, 204...Request unit, 206...Request frequency reduction processing unit, 207...Information processing unit, 301...Information collection device, D1...Input information, D2...Instruction information, D3...Output information, D4...Monitoring information, NW...Public communication network.

Claims

1. An information acquisition unit that acquires input information about the monitored target, A request unit that instructs the AI ​​service to process by transmitting instruction information regarding the input information acquired by the information acquisition unit, The system includes a request frequency reduction processing unit that performs a first determination process to determine whether the process is necessary based on whether a first condition that increases the need for the process is met, When the request frequency reduction processing unit determines that the processing by the generation AI service is necessary, the request unit requests the generation AI service to perform the processing. Information processing system.

2. The aforementioned request frequency reduction processing unit is: If the first decision process determines that the process is necessary, a second decision process is executed to determine whether the process is necessary based on whether a second condition is met, which makes the process more necessary than when the first condition is met. The information processing system according to claim 1.

3. The system includes an information processing unit that processes the input information to improve the judgment accuracy of the AI ​​generation service before transmitting it to the AI ​​generation service. The information processing system according to claim 1.

4. The aforementioned information processing unit Information that would cause a decrease in the accuracy of the processing using the aforementioned generation AI service is removed from the input information. The information processing system according to claim 3.

5. The aforementioned request unit, After requesting processing from the aforementioned generation AI service, the system requests other processing related to the results of that processing from the generation AI service. The information processing system according to claim 1.

6. The aforementioned request unit, This involves requesting two types of generative AI services with different amounts of training data and different numbers of parameters. After commissioning a generative AI service with a small amount of training data and a small number of parameters, The aforementioned request is made to a generative AI service with a large amount of training data and a large number of parameters. The information processing system according to claim 1.

7. The input information is a still image or a moving image of the subject being monitored by the imaging device. The information processing system according to claim 1.

8. In a control method for an information processing system, The first step is to obtain input information about the monitored target, A third step involves sending instruction information regarding the input information obtained in the first step to the generating AI service to instruct it to process, The second step includes executing a first judgment process to determine whether the aforementioned process is necessary based on whether a first condition that increases the need for the aforementioned process is met, If the second step determines that the processing by the generation AI service is necessary, the third step involves requesting the generation AI service to perform the processing. A method for controlling an information processing system.

Images