Monitoring device, monitoring method and program

Abstract

To provide a monitoring device, a monitoring method and a program for reducing unnecessary image processing and image storage.SOLUTION: In a monitoring system 10, a monitoring device has: an acquisition part for acquiring an image of image data picked up by a camera; and a detection part for executing abnormality detection such as an occurrence of visible smoke of a monitoring object from the image when a prescribed characteristic object such as a facility in the image is recognized by using a machine learning model that learns in advance a learning data set composed of a learning image and a correct answer label showing the appropriateness of the prescribed characteristic object in the learning image.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a monitoring device, a monitoring method, and a program. [Background technology]

[0002] In steelworks, exhaust gases emitted from various facilities such as coke ovens and steel converters are treated to meet emission standards before being released. However, visible smoke can sometimes be generated due to furnace damage or unburned gas. From an environmental perspective, it is important to immediately detect the generation of visible smoke and take measures to prevent its generation and recurrence.

[0003] Conventionally, visual monitoring by humans or video surveillance using video recording has been used. However, there are problems such as overlooking visible smoke or checking previous day's monitoring, which can take time from detecting smoke to notifying relevant parties. Furthermore, detecting the location of the smoke source is important for taking measures to prevent visible smoke from occurring. By informing relevant parties of the location and date and time of the smoke occurrence, operational adjustments such as adjusting the flow rate of the smoke-generating equipment or changing the furnace to be operated at the next timing can be made to suppress smoke generation, and trend analysis of smoke generation history can be used to prevent smoke from occurring in the first place. Against this background, smoke detection technology using images captured by cameras has been proposed. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 8-124064 [Patent Document 2] Japanese Patent Application Laid-Open No. 2014-102645 Summary of the Invention [Problem to be solved by the invention]

[0005] When detecting abnormalities such as smoke by image processing, false positives can occur during sunset or rainy weather. A technology is desired that can suppress such false positives and also suppress the accumulation of unnecessary images captured during sunset or rainy weather, when image processing is difficult, thereby reducing the load associated with image processing and image accumulation.

[0006] One object of the present disclosure is to provide techniques for reducing unnecessary image processing and storage. [Means for solving the problem]

[0007] One aspect of the present disclosure relates to a monitoring device having an acquisition unit that acquires an image, and a detection unit that, when a predetermined feature is recognized in the image, performs abnormality detection of a monitored object from the image. [Effects of the Invention]

[0008] According to the present disclosure, unnecessary image processing and image storage can be reduced. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic diagram illustrating anomaly detection according to one embodiment of the present disclosure. [Figure 2] FIG. 2 is a schematic diagram illustrating a captured image according to one embodiment of the present disclosure. [Figure 3] FIG. 3 is a schematic diagram illustrating a monitoring device utilizing a machine learning model according to one embodiment of the present disclosure. [Figure 4] FIG. 4 is a block diagram illustrating a hardware configuration of a monitoring device according to an embodiment of the present disclosure. [Figure 5] FIG. 5 is a block diagram illustrating a functional configuration of a monitoring device according to an embodiment of the present disclosure. [Figure 6] FIG. 6 is a diagram illustrating a feature detection model according to one embodiment of the present disclosure. [Figure 7] FIG. 7 is a diagram illustrating a process for extracting a monitored area image according to an embodiment of the present disclosure. [Figure 8]FIG. 8 is a diagram illustrating a smoke detection model according to one embodiment of the present disclosure. [Figure 9] FIG. 9 is a flowchart illustrating a monitoring process according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0011] In the following embodiment, a monitoring device is disclosed in which the predetermined feature is a predetermined facility and a machine learning model is used to detect anomalies such as the occurrence of visible smoke.

[0012] [Summary of the Disclosure] As shown in FIG. 1 , in a monitoring system 10 according to an embodiment of the present disclosure, a monitoring device 100 acquires image data captured by a camera 30 and determines whether a predetermined feature (e.g., a predetermined facility) is recognized in the acquired image. If the predetermined feature is recognized in the image, the monitoring device 100 performs anomaly detection on the acquired image. Specifically, the monitoring device 100 first determines whether a predetermined feature is captured in an image acquired from the camera 30 using a feature detection model 40 trained to determine whether a predetermined feature is captured in an image. The predetermined feature may be, but is not limited to, a smoke exhaust facility in a steelworks. Next, the monitoring device 100 determines whether visible smoke is generated in the image using a smoke detection model 50 trained to determine whether visible smoke is generated in the image. Note that anomaly detection is not limited to detecting visible smoke generation, and may also include, for example, detecting malfunctions in smoke exhaust facilities.

[0013] Typically, if the camera 30 is not installed to capture the specified smoke exhaust equipment due to installation errors or if the specified smoke exhaust equipment is not captured by the camera 30 due to sunset or rainfall, the monitoring device 100 may not capture the specified smoke exhaust equipment in the image acquired from the camera 30. In this case, even if the monitoring device 100 performs anomaly detection for visible smoke on the image, it may not obtain an appropriate detection result. Therefore, the use of computational resources for anomaly detection and the storage of images captured by the camera 30 may be wasted. To avoid such waste of computational resources and storage space, if the specified smoke exhaust equipment is not captured in the image, the monitoring device 100 may not perform or stop anomaly detection on the image.

[0014] For example, if smoke exhaust facilities 20_1 to 20_3 are set as predetermined facilities, and it is determined that all of the smoke exhaust facilities 20_1 to 20_3 shown in FIG. 2 are recognized, the monitoring device 100 may perform anomaly detection, such as visible smoke generation, on the image. On the other hand, if it is determined that one or more of the smoke exhaust facilities 20_1 to 20_3 are not recognized, the monitoring device 100 may not perform anomaly detection on the image, or may stop any ongoing anomaly detection. In the illustrated embodiment, multiple smoke exhaust facilities 20_1 to 20_3 are set as predetermined features, but the predetermined feature according to the present disclosure is not limited thereto and may be a single feature. Furthermore, as described in detail below, the determination of whether a predetermined feature is recognized in an image may be performed, for example, using a machine learning model, without limitation.

[0015] After acquiring an image of the target for anomaly detection in this manner, the monitoring device 100 performs anomaly detection, such as detecting the occurrence of visible smoke, on the acquired image. For example, as shown in FIG. 3, the monitoring device 100 extracts monitoring area images 21_1 to 21_4 of monitoring areas #1 to #4 from the image. The monitoring device 100 may then acquire detection results indicating the occurrence of visible smoke in each of the acquired monitoring area images 21_1 to 21_4 using smoke detection models #1 to #4 trained for the monitoring areas #1 to #4, respectively. As will be described in detail below, the monitoring device 100 may identify and notify the occurrence of visible smoke and its location based on the detection results of the occurrence of visible smoke in these monitoring areas #1 to #4.

[0016] In this way, the monitoring device 100 detects a specific facility in an image captured by the camera 30, determines whether the image is suitable for anomaly detection, and performs anomaly detection on the image only if the captured image is suitable for anomaly detection. This eliminates the need to perform anomaly detection on images that are not suitable for anomaly detection or to reserve storage space to store images that are not suitable for anomaly detection, thereby avoiding waste of computational and storage resources. Furthermore, when an anomaly such as visible smoke is detected, the monitoring device 100 can identify and notify the facility in which the anomaly was detected, allowing for prompt operational adjustments and / or repairs to be made to the facility.

[0017] Here, the monitoring device 100 may be realized by a computing device such as a server or a personal computer (PC), and may have a hardware configuration such as that shown in Fig. 4. That is, the monitoring device 100 has a storage device 101, a processor 102, an interface device 103, and a communication device 104, which are interconnected via a bus B.

[0018] The programs or instructions that realize the various functions and processes described below in the monitoring device 100 may be downloaded from any external device via a network, or may be provided from a removable storage medium such as a CD-ROM (Compact Disk-Read Only Memory) or flash memory.

[0019] Storage device 101 may be implemented by random access memory, flash memory, a hard disk drive, or the like, and stores installed programs or instructions as well as files, data, etc. used in executing the programs or instructions. Storage device 101 may also include a non-transitory storage medium.

[0020] The processor 102 may be realized by one or more CPUs (Central Processing Units), GPUs (Graphics Processing Units), processing circuitry, etc., which may be composed of one or more processor cores, and performs various functions and processes of the monitoring device 100 described below in accordance with programs, instructions, data such as parameters required to execute the programs or instructions, etc. stored in the storage device 101.

[0021] The interface device 103 realizes an interface with the user of the monitoring device 100. For example, the user operates a GUI (Graphical User Interface) displayed on a display or touch panel using a keyboard, mouse, etc., to send and receive various information, data, instructions, etc. to and from the monitoring device 100 via the interface device 103.

[0022] The communication device 104 is realized by various communication circuits that execute communication processes with external devices, the Internet, a communication network such as a LAN (Local Area Network), and the like.

[0023] However, the above-described hardware configuration is merely an example, and the monitoring device 100 according to the present disclosure may be realized by any other appropriate hardware configuration.

[0024] [Monitoring device] Next, the monitoring device 100 according to an embodiment of the present disclosure will be described with reference to Fig. 5. Fig. 5 is a block diagram showing the functional configuration of the monitoring device 100 according to an embodiment of the present disclosure.

[0025] 5, the monitoring device 100 includes an acquisition unit 110 and a detection unit 120. Each functional unit of the acquisition unit 110 and the detection unit 120 may be realized by the processor 102 executing a computer program stored in the storage device 101 of the monitoring device 100.

[0026] The acquisition unit 110 acquires images. Specifically, the acquisition unit 110 acquires images from a camera 30 that is installed so as to capture an image of a predetermined feature. Typically, the camera 30 may acquire an image every few seconds to every few tens of seconds and transmit the acquired images to the monitoring device 100. Alternatively, the camera 30 may acquire a video or moving image and transmit image frames extracted from the acquired video or moving image every predetermined number of frames to the monitoring device 100. The acquisition unit 110 passes the image data acquired from the camera 30 to the detection unit 120.

[0027] When the detection unit 120 recognizes a predetermined feature in an image, it performs anomaly detection of the monitored object from the image. For example, if the anomaly detection is to detect visible smoke generation, the detection unit 120 may first perform equipment detection and determine whether one or more predetermined smoke exhaust equipment are captured in the image acquired from the acquisition unit 110. If all of the one or more predetermined smoke exhaust equipment are recognized in the image, the detection unit 120 may determine whether visible smoke is being generated for the image. If visible smoke is being generated, the feature 120 may perform a smoke generation notification process according to a predetermined notification rule and notify the visible smoke generation state. On the other hand, if any of the one or more predetermined smoke exhaust equipment are not recognized in the image, the detection unit 120 may not perform a visible smoke generation detection process or may stop any currently running smoke generation detection.

[0028] In this embodiment, the detection unit 120 may determine whether a predetermined feature is recognized in an image by using the feature detection model 40. Specifically, as shown in Fig. 6, the detection unit 120 inputs the image acquired from the acquisition unit 110 to the feature detection model 40, and acquires from the feature detection model 40 a feature detection result indicating whether a predetermined feature is captured in an image.

[0029] For example, the detection unit 120 may start anomaly detection when it continuously acquires a predetermined number of images or a predetermined period of time or more of equipment detection results indicating that all of one or more predetermined pieces of equipment have been recognized. On the other hand, the detection unit 120 may stop ongoing anomaly detection when it continuously acquires a predetermined number of images or a predetermined period of time or more of equipment detection results indicating that any of one or more predetermined pieces of equipment have not been recognized. By utilizing judgments based on the equipment detection results continuously acquired in this manner, the detection unit 120 may be able to avoid erroneous detection due to image noise, erroneous determination by the feature detection model 40, etc.

[0030] Here, the feature detection model 40 may be a machine learning model that has been trained in advance using a training dataset that includes training images and correct labels indicating whether or not a predetermined feature is present in the training images. Typically, the feature detection model 40 can be realized by a neural network that has been trained to receive image data as input and determine whether or not a predetermined feature is recognized in the image data.

[0031] In the learning process, the feature object detection model 40 receives training images from a training dataset as input and outputs a determination result. Then, the parameters of the feature object detection model 40 are adjusted according to the error between the output determination result and the corresponding correct label using the backpropagation algorithm. Once the learning process is complete, the feature object detection model 40 is made available to the detection unit 120. The feature object detection model 40 may be provided within the monitoring device 100, or may be located on an external server or the like that is connected to the monitoring device 100 via communication.

[0032] When it is determined to start anomaly detection based on the above-described equipment detection result, the detection unit 120 may extract a monitoring area image showing a monitoring area set corresponding to the installation positions of one or more predetermined pieces of equipment. For example, as shown in FIG. 7, the detection unit 120 may extract monitoring area images 21_1 to 21_4 corresponding to monitoring areas #1 to #4 from the image. Such monitoring areas #1 to #4 may be set in advance by the operator of the monitoring device 100, for example. The extraction of each monitoring area image can typically be achieved using any known image processing technology depending on the installation positions of the smoke exhaust equipment 20-1 to 20_3.

[0033] In this embodiment, the detection unit 120 may detect smoke from a smoke exhaust facility captured in an image by using the smoke detection model 50. For example, the detection unit 120 may detect smoke in each monitoring area by using the smoke detection model 50 corresponding to each monitoring area. Specifically, when the smoke detection model 50 is trained in advance for each monitoring area and the monitoring area images 21_1 to 21_4 are extracted as described above, the detection unit 120 may detect visible smoke in the monitoring area images 21_1 to 21_4 by using the smoke detection models 50_1 to 50_4 corresponding to the monitoring areas #1 to #4, as shown in FIG. 8.

[0034] For example, the smoke detection models 50_1 to 50_4 may be machine learning models that have been trained in advance using a training data set consisting of training images and ground truth labels indicating the smoke state in the training images. Typically, the smoke detection models 50_1 to 50_4 may be realized by a neural network trained to receive image data as input and determine the smoke state in the image data. Here, the smoke state may indicate whether or not visible smoke is being generated, or may indicate the level of visible smoke (e.g., whether the amount of visible smoke is large, medium, or small). The smoke state may also indicate the color of the visible smoke, the date and time of its generation, etc.

[0035] In the learning process, the smoke detection models 50_1 to 50_4 receive training images of a training data set as input and output a determination result. Then, the parameters of the smoke detection models 50_1 to 50_4 are adjusted according to the error backpropagation method in accordance with the error between the output determination result and the corresponding correct label. Once the learning process is completed, the smoke detection models 50_1 to 50_4 are made available to the detection unit 120. The smoke detection models 50_1 to 50_4 may be provided within the monitoring device 100, or may be located on an external server or the like connected to the monitoring device 100 via communication.

[0036] In one embodiment, the monitoring area includes a notification area, and when smoke is detected in the notification area, the detection unit 120 may execute a smoke notification process. For example, in the specific example described above, monitoring area #4 is set as the notification area. In this case, when the smoke detection model #4 50_4 outputs a detection result of visible smoke generation for the monitoring area image 21_4 of monitoring area #4, the detection unit 120 may start a smoke notification process to notify the occurrence of visible smoke. Typically, the detection unit 120 may notify a terminal of an operator, a related person, or the like that visible smoke has been generated.

[0037] Specifically, when detection results of visible smoke occurrence in the notification area are continuously acquired for a predetermined number of images or for a predetermined period of time or more, detection unit 120 may initiate a smoke notification process. On the other hand, if detection results of visible smoke occurrence in the notification area are continuously acquired for less than the predetermined number of images or for less than the predetermined period of time, detection unit 120 may not execute a smoke notification process. By utilizing a determination based on such continuously acquired detection results of visible smoke occurrence, detection unit 120 may be able to avoid excessive or erroneous notifications.

[0038] In one embodiment, the monitoring area may further include a smoke-producing area corresponding to a smoke exhaust facility. When smoke is detected in both the notification area and the smoke-producing area, the detection unit 120 may execute a smoke notification process to indicate the smoke exhaust facility corresponding to the smoke-producing area. For example, in the specific example described above, monitoring area #4 may be set as the notification area, and monitoring areas #1 to #3 may be set as smoke-producing areas. In this case, when the smoke detection model #4 50_4 outputs a detection result of visible smoke generation for the monitoring area image 21_4 of monitoring area #4, and one of the smoke detection models #1 to #3 50_1 to 50_3 outputs a detection result of visible smoke generation for the monitoring area images #1 to #3 of monitoring areas #1 to #3, the detection unit 120 may start a smoke notification process to notify the operator, relevant parties, etc. of the occurrence of visible smoke and its location. Typically, the detection unit 120 may notify a terminal of an operator, relevant parties, etc., of the occurrence of visible smoke and its location.

[0039] For example, when the smoke detection model #4 50_4 detects the generation of visible smoke in the notification area 21_4 and the smoke detection model #1 50_1 detects the generation of visible smoke in the smoke-generating area 21_1, the detection unit 120 may notify a terminal of an operator or the like that visible smoke has been generated in the smoke-generating area 21_1. Upon receiving such a notification, the operator or the like can recognize that some malfunction has occurred in the operating status or equipment of the smoke exhaust equipment 20_1 corresponding to the smoke-generating area 21_1.

[0040] Here, when detection results of visible smoke occurrence in both the notification area and the smoke-producing area are continuously acquired for a predetermined number of images or for a predetermined period of time or more, the detection unit 120 may start the above-described smoke occurrence notification process. On the other hand, if detection results of visible smoke occurrence in both the notification area and the smoke-producing area are continuously acquired for less than the predetermined number of images or the predetermined period of time, the detection unit 120 may not execute the smoke occurrence notification process. By utilizing judgments based on the continuously acquired detection results of visible smoke occurrence in this way, the detection unit 120 may be able to avoid excessive or erroneous notifications.

[0041] In one embodiment, the detection unit 120 may also identify the operational status of the smoke exhaust equipment 20_1-20_3 at the time of smoke generation. Typically, each piece of equipment using the smoke exhaust equipment 20_1-20_3 operates according to operational information for the kiln or other equipment used in that equipment. For example, the operational information may indicate schedule information, such as the input of materials into the kiln and the kiln temperature, the amount of charged coal, the charging torque, the furnace wall temperature, the soot concentration, and the cock opening. For example, when visible smoke is detected in the monitoring area 21_1 on a certain date and time, the detection unit 120 may identify the operational status of the corresponding smoke exhaust equipment 20_1 on that date and time based on the operational information of the equipment, and report the identified operational status and the visible smoke generation status to an operator or the like. Upon receiving such a report, the operator or the like may analyze the cause of the visible smoke generation from the visible smoke generation status and the operational status, and may adjust operations or repair the equipment.

[0042] [Monitoring process] Next, a monitoring process according to an embodiment of the present disclosure will be described with reference to Fig. 9. The monitoring process is executed by the above-described monitoring device 100, and more specifically, may be realized by one or more processors 102 of the monitoring device 100 executing one or more programs or instructions stored in one or more storage devices 101. Furthermore, the monitoring device 100 may be realized by multiple computers, and the multiple computers may execute the monitoring process. Fig. 9 is a flowchart showing the monitoring process according to an embodiment of the present disclosure.

[0043] 9, in step S101, the monitoring device 100 acquires an image. For example, the monitoring device 100 may acquire an image showing a predetermined facility captured by the camera 30.

[0044] In step S102, the monitoring device 100 determines whether a predetermined feature is captured in the image. Specifically, the monitoring device 100 may determine whether a predetermined feature is recognized in the image by using the feature detection model 40. For example, if all of the predetermined features are recognized in the image, the monitoring device 100 may determine that the predetermined feature is recognized in the image (S102: YES) and proceed to step S103, whereas if any of the predetermined features are not recognized in the image (S102: NO), the monitoring device 100 may determine that the predetermined feature is not recognized in the image and return to step S101.

[0045] To avoid false positives in feature detection, for example, monitoring device 100 may determine whether a predetermined feature is captured in an image based on whether all of the predetermined features are recognized in the images for a predetermined number of consecutive images or for a predetermined period of time or more. Specifically, if all of the predetermined features are captured in the images for a predetermined number of consecutive images or for a predetermined period of time or more, monitoring device 100 may determine that the predetermined feature is recognized in the image. On the other hand, if all of the predetermined features are not recognized in the images for less than the predetermined number of images or for less than the predetermined period of time, monitoring device 100 may determine that the predetermined feature is not recognized in the image.

[0046] In step S103, monitoring device 100 determines whether smoke is being emitted from the smoke exhaust equipment recognized in the image. Specifically, monitoring device 100 may determine whether visible smoke is being recognized in the image using smoke detection model 50. For example, if monitoring device 100 detects visible smoke in the image (S103: YES), it may proceed to step S104, and if monitoring device 100 does not detect visible smoke in the image (S103: NO), it may return to step S101.

[0047] Here, if multiple monitoring areas are set within an image, monitoring device 100 may perform smoke detection for each monitoring area using smoke detection model 50 corresponding to each monitoring area. If monitoring device 100 detects visible smoke in a notification area within a monitoring area, monitoring device 100 may perform smoke notification processing in step S104. On the other hand, if monitoring device 100 does not detect visible smoke in the notification area, monitoring device 100 may acquire the next image in step S101 without proceeding to step S104.

[0048] Furthermore, to avoid false positives or overdetections in smoke detection, for example, monitoring device 100 may determine whether visible smoke is recognized in an image based on whether visible smoke is recognized in the image for a predetermined number of consecutive images or for a predetermined period of time or more. Specifically, if visible smoke is recognized in the image for a predetermined number of consecutive images or for a predetermined period of time or more, monitoring device 100 may determine that visible smoke is recognized in the image. On the other hand, if visible smoke is recognized in the image for less than the predetermined number of images or for less than the predetermined period of time, monitoring device 100 may determine that visible smoke is not recognized in the image.

[0049] In step S104, monitoring device 100 executes a smoke generation notification process. For example, if a notification area is set in the image and visible smoke is detected in the notification area, monitoring device 100 may notify an operator or the like of the occurrence of visible smoke. Furthermore, if visible smoke is detected not only in the notification area but also in the smoke generation area, monitoring device 100 may notify an operator or the like of not only the occurrence of visible smoke but also the location of the smoke.

[0050] In the above-described embodiment, the focus is on the generation of visible smoke as the detection target for abnormality detection, but abnormality detection according to the present disclosure is not necessarily limited to this, and the present disclosure may be applied to detecting any other abnormality that can be detected or recognized by an image.

[0051] According to the above-described embodiment, the monitoring device 100 recognizes predetermined features in an image captured by the camera 30 to determine whether the image is suitable for anomaly detection, and performs anomaly detection on the image only if the captured image is suitable for anomaly detection. This eliminates the need to perform anomaly detection on images that are not suitable for anomaly detection or to reserve memory space to store images that are not suitable for anomaly detection, thereby avoiding waste of computational and memory resources. Furthermore, when an anomaly such as visible smoke is detected, the monitoring device 100 can identify and notify the equipment in which the anomaly was detected, allowing for prompt operational adjustments and / or repairs to be made to the equipment.

[0052] Although the embodiments of the present disclosure have been described in detail above, the present disclosure is not limited to the specific embodiments described above, and various modifications and variations are possible within the scope of the gist of the present disclosure as set forth in the claims. [Explanation of symbols]

[0053] 10. Surveillance System 20 Smoke exhaust equipment 30 Camera 40 Feature detection model 50 Smoke Detection Model 100 Monitoring equipment 110 Acquisition Department 120 Detector

Claims

1. an acquisition unit that acquires an image; a detection unit that detects an abnormality in the monitored object from the image when a predetermined feature is recognized in the image; A monitoring device comprising:

2. The monitoring device according to claim 1 , wherein the detection unit uses a feature detection model to determine whether the predetermined feature is recognized in the image.

3. The monitoring device according to claim 2 , wherein the detection unit stops the abnormality detection when it determines, using the feature detection model, that the predetermined feature is not recognized in the image.

4. The monitoring device according to claim 2 , wherein the feature detection model is trained using a first training data set consisting of training images and correct labels indicating whether the predetermined feature is recognized in the training images.

5. the predetermined feature is a predetermined facility, and the abnormality detected is smoke emission from a smoke exhaust facility; The monitoring device according to claim 1 , wherein the detection unit detects the presence or absence of smoke or the level of smoke emitted from the smoke exhaust equipment captured in the image by using a smoke detection model.

6. The monitoring device of claim 5 , wherein the smoke detection model is trained using a second training data set consisting of training images and ground truth labels indicating smoke states in the training images.

7. The monitoring device according to claim 5 , wherein the detection unit extracts one or more monitoring areas from the image based on the detected position of the predetermined facility.

8. The monitoring device according to claim 5 , wherein the detection unit detects the presence or absence of smoke or the level of smoke in each monitoring area by using a smoke detection model corresponding to each monitoring area.

9. the monitoring area includes a notification area; The monitoring device according to claim 7 , wherein when smoke is detected in the notification area, the detection unit executes a smoke notification process.

10. the monitoring area further includes a smoke generation area corresponding to a smoke exhaust facility; The monitoring device according to claim 9 , wherein when smoke is detected in both the notification area and the smoke-producing area, the detection unit executes a smoke-producing notification process that indicates the smoke exhaust equipment corresponding to the smoke-producing area.

11. The monitoring device according to claim 10 , wherein the detection unit identifies an operating state of the smoke exhaust equipment when smoke is generated.

12. Acquiring an image; When a predetermined feature is recognized in the image, an abnormality detection of the monitored object is performed from the image; The computer performs the monitoring method.

13. Acquiring an image; When a predetermined feature is recognized in the image, an abnormality detection of the monitored object is performed from the image; A program that causes a computer to execute the following.

Images