Water accident detection device, water accident detection system, water accident detection method, and program

Abstract

To provide a technique for appropriately detecting rescue object persons by taking into account their swimming ability.SOLUTION: A water accident detection device includes: an action acquisition unit that acquires action information of swimmers located on or in the water using sensors; a clustering execution unit that divides the swimmers into a plurality of groups; a determination model generation unit that generates a determination model for detecting rescue object persons by using the action information of the swimmers belonging to the groups; and a rescue object person detection unit that detects the rescue object persons by using the determination model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a water disaster detection device, a water disaster detection system, a water disaster detection method, and a program. [Background technology]

[0002] Patent Document 1 discloses a technology related to a drowning determination system. The drowning determination system determines whether a target person A is drowning or is about to drown based on changes in the vertical direction of exposed parts detected in time-series images X. If it is determined that "target person A is drowning or is about to drown," it outputs an alarm signal indicating danger to beach managers, lifeguards, etc. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-061699 Summary of the Invention [Problem to be solved by the invention]

[0004] For example, even if the exposed parts move up and down in the same way, a good swimmer may be swimming, while a poor swimmer may be drowning. Furthermore, even if a poor swimmer does not splash or appear to be drowning, they may still be drowning. The technology disclosed in Patent Document 1 does not take into account factors other than appearance, and therefore may fail to detect a drowning person.

[0005] The present invention has been made in consideration of the above points, and aims to provide a technique for appropriately detecting a person to be rescued in consideration of swimming ability. [Means for solving the problem]

[0006] The present application includes a number of means for solving the above problems, examples of which are as follows.

[0007] In order to solve the above problem, one embodiment of the present invention provides a drowning detection device that includes a behavior acquisition unit that uses a sensor to acquire behavioral information of swimmers located on or underwater, a clustering execution unit that divides the swimmers into multiple groups, a judgment model generation unit that uses the behavioral information of the swimmers belonging to the groups to generate a judgment model for detecting persons to be rescued, and a rescue target detection unit that detects persons to be rescued using the judgment model.

[0008] The judgment model generation unit may generate the judgment model for each group, and the rescue target detection unit may detect the rescue target by identifying the group to which the swimmer's behavioral information belongs.

[0009] The behavior acquisition unit may acquire the behavior information detected by the sensor mounted on the aircraft.

[0010] The behavior acquisition unit may be characterized by acquiring the behavior information detected by multiple sensors, including sensors mounted on an aircraft, sensors fixed on or underwater, and sensors mounted on an underwater moving body.

[0011] The rescue target device may further include a notification generating unit that generates a notification when the rescue target detection unit detects the rescue target.

[0012] The notification may include location information of the person to be rescued.

[0013] The water disaster detection device may include an environmental information acquisition unit that acquires environmental information, which is information about the environment, a rescue timing estimation unit that estimates the rescue timing when rescue will reach the location of the person to be rescued, and a position estimation unit that uses the environmental information to estimate the location of the person to be rescued at the rescue timing, and the notification may be characterized in that it includes location information of the person to be rescued estimated by the position estimation unit.

[0014] The notification may include appearance information of the person to be rescued.

[0015] In addition, in order to solve the above problem, a drowning detection system according to another aspect of the present invention is a drowning detection system including a mobile body and a drowning detection device, wherein the mobile body is equipped with a sensor that detects the behavior of swimmers located on or underwater, and the drowning detection device is characterized by comprising: a behavior acquisition unit that acquires behavior information of swimmers located on or underwater using the sensor; a clustering execution unit that divides the swimmers into multiple groups; a judgment model generation unit that generates a judgment model for detecting persons to be rescued using the behavior information of the swimmers belonging to the groups; and a rescue target detection unit that detects persons to be rescued using the judgment model.

[0016] In addition, in order to solve the above problem, a drowning detection method according to another aspect of the present invention is characterized by comprising a behavior acquisition step of acquiring behavioral information of swimmers located on or underwater using a sensor, a clustering execution step of dividing the swimmers into multiple groups, a judgment model generation step of generating a judgment model for detecting persons to be rescued using the behavioral information of the swimmers belonging to the group, and a rescue target detection step of detecting persons to be rescued using the judgment model.

[0017] In addition, in order to solve the above problem, a program according to another aspect of the present invention is a program that causes a computer processing unit to execute a drowning detection method, and is characterized by executing a behavior acquisition procedure that uses a sensor to acquire behavioral information of swimmers located on or underwater, a clustering execution procedure that divides the swimmers into multiple groups, a judgment model generation procedure that uses the behavioral information of the swimmers belonging to the groups to generate a judgment model that detects persons to be rescued, and a rescue target detection procedure that detects persons to be rescued using the judgment model. [Effects of the Invention]

[0018] According to the present invention, it is possible to appropriately detect a person to be rescued taking into consideration swimming ability.

[0019] Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiments. [Brief explanation of the drawings]

[0020] [Figure 1] FIG. 1 is a diagram showing an example (part 1) of an outline of a water accident detection system. [Figure 2] FIG. 10 is a diagram showing an example (part 2) of an outline of a water accident detection system. [Figure 3] FIG. 2 is a diagram illustrating an example of a functional block of a water accident detection system. [Figure 4] FIG. 4 is a diagram illustrating an example of a data structure of underwater exercise information. [Figure 5] FIG. 2 is a diagram illustrating an example of a hardware configuration of a water accident detection device. [Figure 6] 10 is a flowchart illustrating an example of a behavior information acquisition process. [Figure 7] 10 is a flowchart illustrating an example of a determination model generation process. [Figure 8] 10 is a flowchart illustrating an example of a behavior information monitoring process. [Figure 9] 10 is a flowchart showing an example of a behavior information acquisition process in a first modified example. [Figure 10]10 is a flowchart showing an example of a behavior information monitoring process in the first modified example. [Figure 11] FIG. 10 is a diagram showing an example of functional blocks of a water accident detection system in a second modified example. [Figure 12] 10 is a flowchart showing an example of a behavior information monitoring process in a second modified example. DETAILED DESCRIPTION OF THE INVENTION

[0021] Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a diagram showing an example (part 1) of an outline of a drowning detection system 1. The drowning detection system 1 has a drowning detection device 10 and a plurality of sensors. For example, the drowning detection device 10 is a device owned by a provider of a drowning detection service.

[0022] The drowning detection system 1 uses sensors installed underwater or on the water to acquire behavioral information of the swimmer Y. The sensors include, for example, a fixed camera S that captures images of a beach or a pool, an aircraft T (e.g., a drone), or a mobile camera installed on an underwater mobile object (not shown).

[0023] Swimmer Y with high swimming ability generally swims fast and suppresses up and down body movement while swimming. On the other hand, swimmer Y with low swimming ability generally swims slow, is unstable while swimming, and makes many unnecessary movements. In other words, it is thought that the behavior of swimmers Y with similar levels of swimming proficiency is similar both in and above the water.

[0024] The drowning detection system 1 of this embodiment uses sensors to acquire and cluster the movements of swimmer Y on or underwater, and estimates the level of swimming proficiency by determining which group swimmer Y's behavior belongs to. Furthermore, if swimmer Y's behavior deviates from the behavior corresponding to the swimmer's level of proficiency, the system determines that swimmer Y is a person who needs to be rescued, and outputs a notification to lifeguards, lifeguards, etc.

[0025] FIG. 2 is a diagram showing an example (part 2) of the overview of the drowning detection system 1. In this example, the drowning detection system 1 acquires behavioral information of each swimmer Y using a fixed camera S, which is a sensor installed underwater. The drowning detection system 1 determines the group to which the behavioral information of each swimmer Y belongs and estimates the level of swimming proficiency. For example, the drowning detection system 1 determines that the behavior of swimmer Y2 deviates from the behavior corresponding to the level of proficiency, and outputs a notification.

[0026] 3 is a diagram showing an example of functional blocks of the drowning detection system 1. The drowning detection device 10 is capable of receiving a signal detected by a sensor 21. The drowning detection device 10 is, for example, a server computer or a PC (Personal Computer).

[0027] The drowning detection device 10 includes a processing unit 110, a memory unit 120, an input unit 130, an output unit 140, and a communication unit 150. The processing unit 110 controls the entire drowning detection device 10. The memory unit 120 stores information necessary for processing by the processing unit 110. The input unit 130 accepts information input to the drowning detection device 10 from an input device connected via an input IF 14 (described later). The input unit 130 also accepts input of a signal detected by a sensor 21. The output unit 140 outputs information stored in the drowning detection device 10 from an output device connected via an output IF 15 (described later). The communication unit 150 controls the sending and receiving of information to and from other information processing devices connected for communication.

[0028] The processing unit 110 includes a behavior acquisition unit 111, a swimmer determination unit 112, a clustering execution unit 113, a determination model generation unit 114, a rescue target detection unit 115, a notification generation unit 116, and a mobile object control unit 117. The behavior acquisition unit 111 acquires behavior information of a swimmer Y positioned on or underwater using a sensor 21. The behavior information includes, for example, information indicating the posture of the swimmer Y, skeletal information, movement information of parts of the skeleton, acceleration, angular velocity, etc. As an example, the behavior acquisition unit 111 acquires behavior information obtained by vectorizing the behavior of the swimmer Y from an image acquired by the sensor 21 using a known technique such as optical flow.

[0029] The behavior acquisition unit 111 may acquire behavior information detected by any of the sensors 21 mounted on the aircraft T, the sensors 21 mounted on an underwater moving body, or the sensors 21 fixed on or underwater. The behavior acquisition unit 111 may acquire behavior information detected by multiple sensors 21 out of these sensors 21.

[0030] The swimmer determination unit 112 identifies the swimmer Y using information acquired by the sensor 21. For example, the swimmer determination unit 112 identifies the swimmer Y by applying a known face recognition technique to a facial image of the swimmer Y captured by a camera, which is the sensor 21. Furthermore, for example, the swimmer determination unit 112 identifies the swimmer Y based on the body shape, swimsuit, etc. of the swimmer Y captured by the camera. Note that when multiple swimmers Y are captured in the image, the swimmer determination unit 112 identifies each of the swimmers Y.

[0031] The clustering execution unit 113 divides the swimmers Y into multiple groups based on the behavioral information. The clustering method is not limited, and any of non-hierarchical clustering such as the k-means method, and hierarchical clustering such as the group average method, Ward's method, shortest distance method, or longest distance method may be used. The clustering execution unit 113 extracts features from the behavioral information to cluster the swimmers Y who have similar behaviors so that they belong to the same group. As a result, the swimmers Y are divided into groups according to their level of swimming proficiency.

[0032] The determination model generation unit 114 generates a determination model for detecting rescue targets using behavioral information belonging to the group. The determination model generation unit 114 performs machine learning using behavioral information of multiple swimmers Y belonging to the group, thereby generating, for each group, a determination model that indicates a normal range of behavioral information for the group. For example, the determination model generation unit 114 can generate a determination model using known techniques such as multiple regression analysis, decision tree, logistic regression, and random forest.

[0033] The rescue target person detection unit 115 detects rescue targets using a judgment model. It can be said that the rescue target person detection unit 115 estimates the swimmer Y's swimming proficiency level by identifying the group to which the swimmer Y's behavior information belongs. The rescue target person detection unit 115 detects rescue targets using a judgment model generated for the group to which the swimmer Y belongs. The rescue target person detection unit 115 judges that the swimmer Y is a rescue target when the swimmer Y's behavior information deviates from the normal range of behavior information for the group to which the swimmer Y belongs.

[0034] When the rescue target person detection unit 115 detects a rescue target person, the notification generation unit 116 identifies the position of the rescue target person and generates a notification including the position information. The mobile object control unit 117 controls the movement of the mobile object when the sensor 21 is mounted on a mobile object including an air vehicle T. For example, the mobile object control unit 117 controls the mobile object to perform a predetermined movement during a predetermined time period.

[0035] The storage unit 120 stores underwater exercise information 121. The underwater exercise information 121 is information in which each swimmer Y identified by the swimmer determination unit 112 is associated with behavior information.

[0036] The sensor 21 detects behavioral information of the swimmer Y. In this embodiment, an example will be described in which the sensor 21 is a camera, but the sensor 21 is not limited to a camera as long as it can detect behavioral information of the swimmer Y in a non-contact manner. For example, any sensor that detects the behavior of the swimmer Y, such as an inertial sensor, a gyro sensor, or a geomagnetic sensor, can be used.

[0037] FIG. 4 is a diagram showing an example of the data structure of the underwater exercise information 121. For example, the underwater exercise information 121 includes a personal ID and behavior information. The personal ID is identification information that specifies the swimmer Y identified by the swimmer determination unit 112. The behavior information is behavior information of the swimmer Y specified by the personal ID. For example, the underwater exercise information 121 has behavior information acquired by one sensor 21 as one record. That is, in this example, when multiple sensors 21 are used simultaneously, multiple records are generated for one swimmer Y.

[0038] 5 is a diagram showing an example of the hardware configuration of the drowning detection device 10. The drowning detection device 10 includes a calculation device 11, a memory 12, an external storage device 13, an input IF (Interface) 14, an output IF 15, and a communication IF 16, and each component is connected by a bus.

[0039] The arithmetic unit 11 is a calculation device such as a CPU (Central Processing Unit), and executes processing according to a program recorded in the memory 12 or the external storage device 13. In the drowning detection device 10, processing is performed by the arithmetic unit 11 that operates according to a program read onto the memory 12 or the external storage device 13. The processing unit 110 realizes each function by the arithmetic unit 11 executing the program.

[0040] The memory 12 is a storage device such as RAM (Random Access Memory) or flash memory, and functions as a storage area from which programs and data are temporarily read. The external storage device 13 is a writable and readable storage medium and storage media drive, such as an HDD (Hard Disk Drive), CD-R (Compact Disc-Recordable), or DVD-RAM (Digital Versatile Disk-Random Access Memory). The functions of the storage unit 120 are realized by the memory 12 or the external storage device 13. Note that the functions of the storage unit 120 may also be realized by a storage device connected via the communication IF 16.

[0041] The input IF 14 is an interface for receiving input operations from an operator, and is connected to input devices such as a touch panel, keyboard, mouse, microphone, etc. The output IF 15 is an interface for outputting information to an output device such as an OLED (Organic Light Emitting Diode) display built into the drowning detection device 10.

[0042] The communication IF 16 is an interface for connecting the drowning detection device 10 to a network, and is connected to a communication device such as a LAN (Local Area Network) card. The drowning detection device 10 may also have a storage medium drive (not shown) for inputting and outputting information from portable media such as a CD (Compact Disk) or a DVD (Digital Versatile Disk).

[0043] The processing of each component of the drowning detection device 10 may be executed by one piece of hardware or by multiple pieces of hardware. Furthermore, the processing of each component of the drowning detection device 10 may be realized by one program or by multiple programs.

[0044] FIG. 6 is a flowchart showing an example of behavioral information acquisition processing. This processing is executed to extract behavioral information of swimmer Y from information detected by a sensor. In the example shown below, sensor 21 is a fixed camera S or a mobile camera installed on a mobile body such as an aircraft T. Before starting this processing, sensor 21 starts capturing images and transmits the captured images to drowning detection device 10 while capturing images. This flowchart starts, for example, when the behavior acquisition unit 111 of drowning detection device 10 acquires images captured by sensor 21. Note that if drowning detection device 10 simultaneously acquires images from multiple sensors 21, the processing of this flowchart is executed for each image.

[0045] First, the swimmer determination unit 112 determines whether or not personal identification information has been detected (step S11). Specifically, the swimmer determination unit 112 determines whether or not the image acquired by the behavior acquisition unit 111 includes information that can identify an individual, such as a face image, body type, swimsuit, etc. If the swimmer determination unit 112 determines that personal identification information has not been detected ("NO" in step S11), it repeats this process until personal identification information is detected. Note that if multiple swimmers Y appear in the image, the swimmer determination unit 112 performs the following process for each swimmer Y.

[0046] When the swimmer determination unit 112 determines that personal identification information has been detected ("YES" in step S11), it generates a new record in the underwater exercise information 121 (step S12). Specifically, the swimmer determination unit 112 generates a new personal ID and includes it in the new record in the underwater exercise information 121.

[0047] Next, the behavior acquisition unit 111 acquires behavior information (step S13). Specifically, the behavior acquisition unit 111 extracts the behavior information of swimmer Y from the image using a known method and includes it in the record of underwater exercise information 121 generated in step S12.

[0048] Next, the swimmer determination unit 112 determines whether the personal identification information has been lost (step S14). The swimmer determination unit 112 determines whether the personal identification information detected in step S11 has been lost from the image. Note that the case where the personal identification information has been lost from the image means, for example, that swimmer Y has left the monitoring range of the sensor by getting out of the pool. If the swimmer determination unit 112 determines that the personal identification information has not been lost ("NO" in step S14), the behavior acquisition unit 111 continues the processing of step S13. If the swimmer determination unit 112 determines that the personal identification information has been lost ("YES" in step S14), the processing unit 110 ends the processing of this flowchart.

[0049] FIG. 7 is a flowchart showing an example of a judgment model generation process. This process is executed as a preparation stage for detecting a person to be rescued in the behavior information monitoring process (FIG. 8) described below. The process of this flowchart is started, for example, when an input operation of a judgment model generation command is accepted. Note that the process of this flowchart may be started, for example, periodically, in the drowning detection device 10.

[0050] First, the clustering execution unit 113 receives a designation of a distance definition (step S21). Specifically, the clustering execution unit 113 receives a designation, via an input operation by the user, of a distance definition for determining whether or not each piece of behavioral information is similar. Note that the definition of distance used for clustering may be a pre-designated definition.

[0051] Next, the clustering execution unit 113 clusters the behavior information included in the underwater exercise information 121 based on the feature amounts (step S22). For example, the clustering execution unit 113 generates a feature vector having a predetermined number of components from the time-series behavior information included in the underwater exercise information 121, and clusters the information based on the distance specified in step S21. Before clustering, the clustering execution unit 113 may extract part of the behavior information based on the feature amounts. For example, the clustering execution unit 113 clusters the behavior of swimmer Y using swimmer Y's stopping time, moving time, moving speed, vertical movement of the body, moving speed of each part, etc.

[0052] Note that the pre-processing performed when clustering the behavioral information is not limited to the above example, and as for the clustering method, any existing method can be used, as described above.

[0053] Next, the determination model generation unit 114 generates a determination model for detecting rescue targets for each group (step S23). As an example, the determination model generation unit 114 performs machine learning using a plurality of pieces of behavioral information determined to belong to the same group as a result of the clustering in step S22, and generates a determination model that indicates a normal range. The determination model generation unit 114 stores the generated determination model in the storage unit 120. Thereafter, the processing unit 110 ends the processing of this flowchart.

[0054] 8 is a flowchart showing an example of behavior information monitoring processing. This processing is executed to detect rescue targets in real time from information acquired by sensors. Before starting this processing, the drowning detection device 10 has completed the determination model generation processing shown in FIG. 7. The drowning detection unit of the drowning detection device 10 executes this processing for each piece of behavior information that continues to be acquired in step S13 of the behavior information acquisition processing shown in FIG. 6.

[0055] First, the rescue target detection unit 115 identifies behavioral information to be processed (step S31). Specifically, the rescue target detection unit 115 identifies one of the behavioral information acquired in step S13 from the information detected by the sensor as the processing target.

[0056] Next, the rescue target detection unit 115 determines a group to which the behavior information belongs (step S32). Specifically, the rescue target detection unit 115 determines a group to which at least a part of the behavior information identified as the processing target in step S31 belongs, using one or more judgment models generated in the judgment model generation process. For example, when the behavior information over time belongs to the normal range of multiple groups, the rescue target detection unit 115 determines the group that has been in the normal range for the longest time as the group to which the behavior information belongs. In this way, it can be said that the rescue target detection unit 115 estimates the swimming proficiency level of swimmer Y associated with the behavior information to be processed.

[0057] Next, the rescue target person detection unit 115 determines whether or not a rescue target person has been detected (step S33). Specifically, the rescue target person detection unit 115 identifies the judgment model generated in the judgment model generation process of FIG. 7 for the group determined in step S32. The rescue target person detection unit 115 uses the judgment model to determine whether or not behavioral information from a predetermined timing to the present time, among the behavioral information of the processing target, is within a normal range. For example, the rescue target person detection unit 115 determines whether or not behavioral information from one minute before to the present time is within a normal range. If the rescue target person detection unit 115 determines that the behavioral information of the processing target is not within the normal range, it considers that a rescue target person has been detected.

[0058] When determining that a rescue target person has not been detected ("NO" in step S33), the rescue target person detection unit 115 determines whether the end of the behavioral information has been detected (step S34). Specifically, when determining that personal identification information has been lost in step S14 of Fig. 6 for the behavioral information to be processed, the rescue target person detection unit 115 regards the end of the behavioral information as having been detected.

[0059] If the rescue target person detection unit 115 determines that the end of the behavior information has not been detected ("NO" in step S34), the rescue target person detection unit 115 proceeds to step S33. That is, in step S33, the rescue target person detection unit 115 continues to monitor the behavior information.

[0060] When the rescue target person detection unit 115 determines that the end of the behavioral information has been detected ("YES" in step S34), it determines whether or not there is unprocessed behavioral information (step S35). Specifically, the rescue target person detection unit 115 determines whether or not there is behavioral information that has not yet been determined as a processing target in step S31 among the behavioral information that continues to be acquired in step S13 of Fig. 6. When the rescue target person detection unit 115 determines that there is no unprocessed behavioral information ("NO" in step S35), the processing unit 110 ends the processing of this flowchart.

[0061] When the rescue target detection unit 115 determines that unprocessed behavior information exists (YES in step S35), the process proceeds to step S31. That is, in step S31, the rescue target detection unit 115 identifies one piece of behavior information from the unprocessed behavior information as a processing target.

[0062] When the rescue target person detection unit 115 determines that a rescue target person has been detected ("YES" in step S33), the notification generation unit 116 outputs a notification including information about the rescue target person (step S36). For example, the notification generation unit 116 generates a notification including location information and appearance information about the rescue target person, and displays the notification on a display connected via the output IF 15. Note that the notification generation unit 116 may output the notification by transmitting the notification to a terminal device (not shown) owned by a lifeguard or the like.

[0063] In this example, the notification generation unit 116 identifies the location information of the person to be rescued using the location of the sensor 21 that acquired the behavioral information of the processing target, and includes the location information in the notification. If the sensor 21 is fixed, the location of the sensor 21 is stored in advance in the storage unit 120. If the sensor 21 is attached to a mobile object including the flying object T, the notification generation unit 116 acquires the location of the mobile object by receiving positioning signals from GPS (Global Positioning System) satellites, quasi-zenith satellite systems, etc., and identifies the location of the sensor 21. Furthermore, the notification generation unit 116 identifies an image of the person to be processed using the behavioral information of the processing target, and includes the image in the notification as appearance information.

[0064] The process executed in step S37 thereafter is the same as the process executed in step S35, and therefore the description thereof will be omitted. Then, the processing unit 110 ends the process of this flowchart.

[0065] As described above, according to this embodiment, whether swimmer Y is a person to be rescued can be determined based on the swimmer's level of proficiency in swimming, and therefore, rescue persons can be appropriately detected. Furthermore, since the output notification includes information about the person to be rescued, managers of swimming facilities and lifeguards can quickly carry out rescue operations.

[0066] <First Modification> Next, a drowning detection system 1 in a first modified example will be described. Differences from the above-described embodiment will be described below. In the first modified example, when behavioral information detected by different sensors 21 is behavioral information of the same person to be rescued, the drowning detection device 10 associates the behavioral information with each other, thereby preventing notifications from being output for each of the behavioral information.

[0067] Fig. 9 is a flowchart showing an example of behavioral information acquisition processing in Modification 1. The processing of this flowchart starts, for example, when the behavior acquisition unit 111 of the drowning detection device 10 acquires an image captured by the sensor 21, similar to the behavioral information acquisition processing of Fig. 6.

[0068] The processes executed in steps S11 and S12 are the same as those executed in steps S11 and S12 in Fig. 6, and therefore description thereof will be omitted. In step S12, the swimmer determination unit 112 associates the personal identification information detected in step S11 with a new record of the underwater exercise information 121.

[0069] Next, the swimmer determination unit 112 determines whether or not personal identification information of the same swimmer Y has been acquired by another sensor 21 (step S121). Specifically, the swimmer determination unit 112 refers to personal identification information associated with behavior information already acquired by another sensor 21, and determines whether or not personal identification information common to the personal identification information detected in step S11 has been acquired by another sensor 21.

[0070] If the swimmer determination unit 112 determines that the personal identification information of the same swimmer Y has not been acquired by another sensor 21 ("NO" in step S121), the process proceeds to step S13.

[0071] If the swimmer determination unit 112 determines that the personal identification information of the same swimmer Y has been acquired by another sensor 21 ("YES" in step S121), it associates the record of the target underwater exercise information 121 (step S122). Specifically, the swimmer determination unit 112 associates the record of the underwater exercise information 121 of the same swimmer Y, which was determined to have been acquired by another sensor 21 in step S121, with the record of the underwater exercise information 121 newly generated in step S12.

[0072] Thereafter, the processes performed in steps S13 and S14 are the same as those performed in steps S13 and S14 in FIG. 6, and therefore description thereof will be omitted.

[0073] Fig. 10 is a flowchart showing an example of behavior information monitoring processing in Modification 1. The processing of this flowchart is executed for each piece of behavior information that continues to be acquired in step S13 of the behavior information acquisition processing shown in Fig. 6, similar to the behavior information monitoring processing shown in Fig. 8.

[0074] The processing performed in steps S31 to S35 is the same as the processing performed in steps S31 to S35 shown in FIG. 8, and therefore a description thereof will be omitted.

[0075] When it is determined in step S33 that a rescue target has been detected ("YES" in step S33), the rescue target detection unit 115 determines whether or not other related behavior information exists (step S331). Specifically, the rescue target detection unit 115 determines whether or not other behavior information exists associated with the behavior information to be processed, i.e., the record of the underwater exercise information 121, in step S122 of Fig. 9. If the behavior information to be processed is associated with other behavior information, it means that behavior information of one rescue target has been acquired by multiple sensors 21.

[0076] If the rescue target detection unit 115 determines that there is no other related behavior information (NO in step S331), the process proceeds to step S36.

[0077] When it is determined that other related behavior information exists (YES in step S331), the rescue target detection unit 115 determines whether a notification has already been output (step S332). Specifically, the rescue target detection unit 115 determines whether a notification has already been output in step S36 in the behavior information monitoring process performed on the other behavior information determined to be related to the behavior information to be processed in step S331.

[0078] If the rescue target detection unit 115 determines that a notification has not been output ("NO" in step S332), the process proceeds to step S36. The processes executed in steps S36 to S37 are the same as the processes executed in steps S36 to S37 in Fig. 8, and therefore description thereof will be omitted. Note that in step S36, the notification generation unit 116 associates information indicating that a notification has been output with the record of the underwater exercise information 121 related to the behavior information of the processing target for which the notification has been output.

[0079] If it is determined that the notification has already been output (YES in step S332), the rescue target detection unit 115 executes the process of step S38. The process of step S38 is the same as step S35, and therefore the description thereof will be omitted.

[0080] As described above, according to this modified example, when behavioral information of swimmer Y is obtained using multiple sensors 21, if multiple sensors 21 determine that swimmer Y is a person to be rescued, it is possible to prevent duplicate notifications from being output.

[0081] <Second Modification> Next, a drowning detection system 1 in a second modified example will be described. Differences from the above-described embodiment will be described below. The drowning detection device 10 in the second modified example estimates the position of the person to be rescued using environmental information and includes the estimated position in the notification.

[0082] 11 is a diagram showing an example of functional blocks of a drowning detection system 1 in the second modified example. The processing unit 110 of the drowning detection device 10 in the second modified example includes a behavior acquisition unit 111, a swimmer determination unit 112, a clustering execution unit 113, a determination model generation unit 114, a rescue target detection unit 115, a notification generation unit 116, a mobile object control unit 117, an environmental information acquisition unit 1111, a rescue timing estimation unit 1112, and a position estimation unit 1113.

[0083] The environmental information acquisition unit 1111 acquires environmental information, which is information about the environment. For example, the environmental information acquisition unit 1111 acquires information such as tides, tidal currents, wind speed, and weather. The environmental information acquisition unit 1111 may acquire environmental information detected by the sensor 21 or another sensor (not shown), or may acquire environmental information from another device communicably connected via a network (not shown). The rescue timing estimation unit 1112 estimates the rescue timing when rescue will arrive at the position of the person to be rescued. For example, the rescue timing estimation unit 1112 estimates the rescue timing when a rescuer moving at a predetermined speed will arrive at the position of the person to be rescued using a known method.

[0084] The position estimation unit 1113 estimates the position of the rescue target at the time of rescue using environmental information. The method by which the position estimation unit 1113 estimates the position is not limited, and known methods can be used. For example, the position estimation unit 1113 may estimate the position of the rescue target by generating a position estimation model for estimating the position of the swimmer Y by performing machine learning using behavioral information of multiple swimmers Y and environmental information at the time the behavioral information was acquired. As one example, in the case where a swimmer Y who drowned while swimming in the ocean gradually moves along with the current, the position estimation unit 1113 estimates the position of the swimmer Y at the time a lifeguard arrives. As another example, the position estimation unit 1113 estimates the position of a swimmer Y who drowned in a lazy river at the time a lifeguard arrives.

[0085] Fig. 12 is a flowchart showing an example of behavior information monitoring processing in Modification 2. The processing of this flowchart is executed for each piece of behavior information that continues to be acquired in step S13 of the behavior information acquisition processing shown in Fig. 6, similar to the behavior information monitoring processing shown in Fig. 8.

[0086] The processing performed in steps S31 to S35 is the same as the processing performed in steps S31 to S35 shown in FIG. 8, and therefore a description thereof will be omitted.

[0087] If the rescue target person detection unit 115 determines in step S33 that a rescue target person has been detected (YES in step S33), the environmental information acquisition unit 1111 acquires environmental information (step S333).

[0088] Next, the position estimation unit 1113 estimates the position of the rescue target at the time when rescue will arrive (step S334). Specifically, the rescue timing estimation unit 1112 estimates the rescue timing using, for example, the moving speed of the rescuer. The position estimation unit 1113 estimates the position of the rescue target at the rescue timing estimated by the rescue timing estimation unit 1112 using the environmental information acquired in step S333.

[0089] Next, the notification generation unit 116 outputs a notification including the position information of the person to be rescued at the estimated time (step S335). Specifically, the notification generation unit 116 generates a notification including the position of the person to be rescued estimated in step S334, and displays the notification on a display connected via the output IF 15. The notification may include appearance information of the person to be rescued.

[0090] The process performed in step S37 is the same as the process performed in step S37 shown in FIG. 8, and therefore a description thereof will be omitted.

[0091] As described above, according to this modification, the location of the person to be rescued can be estimated taking into account the influence of environmental information and can be included in the notification, thereby enabling the rescuer to reach the location of the person to be rescued more quickly.

[0092] Although the above describes each embodiment of the present invention, the present invention is not limited to the above-described exemplary embodiment and includes various modifications. For example, the above-described exemplary embodiment has been described in detail to facilitate understanding of the present invention, and the present invention is not limited to an embodiment including all of the components described herein. Furthermore, part of the components of one exemplary embodiment can be replaced with the components of another exemplary embodiment. Furthermore, the components of another exemplary embodiment can be added to the components of one exemplary embodiment. Furthermore, part of the components of each exemplary embodiment can be added, deleted, or replaced with other components. Furthermore, some or all of the above-described components, functions, processing units, processing means, etc. may be implemented in hardware, for example, by designing them as integrated circuits. Furthermore, the control lines and information lines in the figures are only those considered necessary for explanation, and not necessarily all of them are shown. It can be assumed that almost all components are interconnected.

[0093] Furthermore, the functional configuration of the drowning detection device 10 described above has been classified according to the main processing content for ease of understanding. The classification method and names of the components do not limit the present invention. As described above, the configuration of the drowning detection device 10 can be classified into more components according to the processing content. Furthermore, it is also possible to classify each component so that it performs even more processing. [Explanation of symbols]

[0094] 1: drowning detection system, 10: drowning detection device, 11: arithmetic unit, 12: memory, 13: external storage device, 14: input IF, 15: output IF, 16: communication IF, 21: sensor, 110: processing unit, 111: behavior acquisition unit, 112: swimmer determination unit, 113: clustering execution unit, 114: determination model generation unit, 115: rescue target person detection unit, 116: notification generation unit, 117: mobile object control unit, 120: memory unit, 121: underwater motion information, 130: input unit, 140: output unit, 150: communication unit, 1111: environmental information acquisition unit, 1112: rescue timing estimation unit, 1113: position estimation unit, S: fixed camera, T: flying object, Y: swimmer

Claims

1. a behavior acquisition unit that acquires behavior information of a swimmer located on or underwater using a sensor; a clustering execution unit that divides the swimmers into a plurality of groups; a determination model generation unit that generates a determination model for detecting a person to be rescued using behavior information of the swimmers belonging to the group; A water accident detection device comprising: a rescue target detection unit that detects a rescue target using the judgment model.

2. The water accident detection device according to claim 1, the determination model generation unit generates the determination model for each of the groups; A drowning detection device, characterized in that the rescue target detection unit detects the rescue target by identifying the group to which the swimmer's behavior information belongs.

3. The water accident detection device according to claim 1, A drowning detection device characterized in that the behavior acquisition unit acquires the behavior information detected by the sensor mounted on the aircraft.

4. The water accident detection device according to claim 1, A water disaster detection device characterized in that the behavior acquisition unit acquires the behavior information detected by multiple sensors, including sensors mounted on an aircraft, sensors fixed on or underwater, and sensors mounted on an underwater moving body.

5. The water accident detection device according to claim 1, A water disaster detection device comprising a notification generation unit that generates a notification when the rescue target person detection unit detects the rescue target person.

6. The water accident detection device according to claim 5, A water disaster detection device, characterized in that the notification includes location information of the person to be rescued.

7. The water accident detection device according to claim 5 or 6, an environmental information acquisition unit that acquires environmental information that is information related to the environment; a rescue timing estimation unit that estimates a rescue timing when rescue will reach the position of the person to be rescued; a position estimation unit that estimates a position of the person to be rescued at the rescue timing by using the environmental information, A water disaster detection device, characterized in that the notification includes location information of the person to be rescued estimated by the location estimation unit.

8. The water accident detection device according to claim 5 or 6, A water disaster detection device, characterized in that the notification includes appearance information of the person to be rescued.

9. A water disaster detection system including a moving body and a water disaster detection device, The moving body is A sensor is provided to detect the behavior of a swimmer located on or underwater, The water detection device is a behavior acquisition unit that acquires behavior information of swimmers located on or underwater using the sensor; and a clustering execution unit that classifies the swimmers into a plurality of groups. a determination model generation unit that generates a determination model for detecting a person to be rescued using behavior information of the swimmers belonging to the group; A water accident detection system comprising: a rescue target person detection unit that detects a rescue target person using the judgment model.

10. a behavior acquisition step of acquiring behavior information of a swimmer located on or underwater using a sensor; a clustering step for dividing the swimmers into a plurality of groups; a determination model generation step of generating a determination model for detecting a person to be rescued using behavior information of the swimmers belonging to the group; A water accident detection method comprising a rescue target detection procedure for detecting a rescue target using the judgment model.

11. A program that causes a processing unit of a computer to execute a drowning detection method, a behavior acquisition step of acquiring behavior information of a swimmer located on or underwater using a sensor; a clustering step for dividing the swimmers into a plurality of groups; a determination model generation step of generating a determination model for detecting a person to be rescued using behavior information of the swimmers belonging to the group; and a rescue target person detection procedure for detecting a rescue target person using the judgment model.

Images