Monitoring device, monitoring method, and program

1 . A monitoring device comprising:

a camera that repeatedly captures a captured image of a monitoring target;

a processor configured to:

determine a type of the monitoring target included in the captured image captured by the camera by applying the captured image to a learning device for image classification;

store a plurality of pieces of correspondence information for associating a type of the monitoring target with one or more types of abnormality of a detection target;

specify one or more types of abnormality of the detection target corresponding to the determined type of the monitoring target using the correspondence information;

detect one or more abnormalities by applying the captured image captured by the camera to one or more monitoring models that are respectively used to detect the specified one or more types of abnormality, each monitoring model being configured to detect a single type of abnormality related to the monitoring target included in the captured image,

wherein each monitoring model is a learning device learned using a plurality of sets of training input information that is a captured image and training output information indicating presence or absence of an abnormality related to a monitoring target included in the captured image of the training input information; and

output, when the abnormality is detected an output related to detection of the abnormality; and

a moving platform operatively connected to the monitoring device to move the monitoring device.

2 . The monitoring device according to claim 1 , wherein the processor is configured to acquire one or more monitoring models corresponding to the type of the monitoring target from a server that holds the one or more monitoring models, wherein the processor is further configured to detect one or more abnormalities using the acquired one or more monitoring models.

3 . The monitoring device according to claim 1 , wherein the processor is configured to, when it is determined that a plurality of the types of the monitoring targets are included in the captured image, detect one or more abnormalities using a plurality of monitoring models respectively corresponding to the plurality of types of monitoring targets that are determination results.

4 . The monitoring device according to claim 3 , wherein the processor is configured to when it is determined that a plurality of the types of the monitoring targets are included in the captured image, detect, for each part of the captured image corresponding to each of the types of the monitoring targets that are determination results, a respective abnormality using a respective monitoring model corresponding to the type of the monitoring target and the respective abnormality.

5 . The monitoring device according to claim 1 , wherein the processor is further configured to output according to a certainty factor corresponding to the one or more abnormalities detected.

6 . The monitoring device according to claim 1 , wherein

the captured image also includes sound, and

the processor is configured to detect the one or more abnormalities by also using the sound included in the captured image.

7 . A monitoring method comprising:

capturing repeatedly, by a camera, a captured image of a monitoring target;

determining repeatedly a type of a monitoring target included in the captured image captured in the step of capturing the captured image by applying the captured image to a learning device for image classification;

specifying one or more types of abnormality of a detection target corresponding to the determined type of the monitoring target using correspondence information, a plurality of pieces of correspondence information for associating a type of the monitoring target with one or more types of abnormality of a detection target being stored in a correspondence information storage;

detecting one or more abnormalities by applying the captured image captured in the step of capturing the captured image to one or more a monitoring models that are respectively used to detect the specified one or more types of abnormality, correspondence information with the type of the monitoring target determined in the step each monitoring model being configured to detect a single type of abnormality related to the monitoring target included in the captured image,

wherein each monitoring model is a learning device learned using a plurality of sets of training input information that is a captured image and training output information indicating presence or absence of an abnormality related to a monitoring target included in the captured image of the training input information;

a step of, when the abnormality is detected in the step of detecting the abnormality, performing an output related to detection of the abnormality; and

moving the camera by a moving platform.

8 . A computer program product comprising a non-transitory computer-readable medium that when executed by a processor causes a computer to execute:

a step of repeatedly determining a type of a monitoring target included in a captured image of the monitoring target by applying the captured image to a learning device for image classification, wherein the captured image is repeatedly captured by a camera;

a step of specifying one or more types of abnormality of a detection target corresponding to the determined type of the monitoring target using correspondence information, a plurality of pieces of correspondence information for associating a type of the monitoring target with one or more types of abnormality of a detection target being stored in a correspondence information storage;

a step of detecting one or more abnormalities by applying the captured image of the monitoring target to one or more monitoring models that are respectively used to detect the specified one or more types of abnormality each monitoring model being configured to detect a single type of abnormality related to the monitoring target included in the captured image,

wherein each monitoring model is a learning device learned using a plurality of sets of training input information that is a captured image and training output information indicating presence or absence of an abnormality related to a monitoring target included in the captured image of the training input information;

a step of, when the abnormality is detected in the step of detecting the abnormality, performing an output related to detection of the abnormality, and

a step of moving the camera by controlling a moving platform.