Detection system, detection method, and program

Abstract

To improve the accuracy of smoke and flame detection in waste pits. [Solution] Based on estimation image data showing a captured image of the waste pit 110 where waste is stored, and a trained model, the system detects the generation of at least one of smoke and flames in the waste pit 110, and also detects the generation of at least one of smoke and flames in the waste pit 110 using an invisible light sensor.

Description

Technical Field

[0001] The present disclosure relates to a detection system, a detection method, and a program.

Background Art

[0002] It is desirable to detect smoke and flames generated from waste stored in a garbage pit at an early stage. As a technique for detecting smoke and flames of waste stored in a garbage pit, there is a technique described in Patent Document 1.

[0003] Patent Document 1 discloses detecting a fire point or a smoke generation point based on an infrared image and a visible light image captured using at least one pair of infrared / visible composite cameras.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in Patent Document 1, detection of smoke and flames is performed only based on the result of image processing. Therefore, it is not easy to accurately detect smoke and flames in a garbage pit with the technique described in Patent Document 1.

[0006] The present disclosure has been made in view of the above problems, and an object thereof is to improve the detection accuracy of smoke and flames in a garbage pit.

Means for Solving the Problems

[0007] The detection system of this disclosure is characterized by comprising: an acquisition means for acquiring estimation image data showing a photographic image of a waste pit where waste is stored; a trained model that receives the photographic image showing the photographic image and estimates the occurrence of at least one of smoke and flames in the waste pit; a first detection means for detecting the occurrence of at least one of smoke and flames based on the estimation image data acquired by the acquisition means; and a second detection means for detecting the occurrence of at least one of smoke and flames using an invisible light sensor.

[0008] The detection method of this disclosure is characterized by comprising: an acquisition step of acquiring estimation image data showing a photographic image of a waste pit where waste is stored; a first detection step of detecting the occurrence of at least one of smoke and flames based on a trained model that receives the said photographic image data and estimates the occurrence of at least one of smoke and flames in the waste pit, and the estimation image data acquired by the acquisition step; and a second detection step of detecting the occurrence of at least one of smoke and flames using an invisible light sensor.

[0009] The program of this disclosure is for causing a computer to function as one of the means of the detection system. [Effects of the Invention]

[0010] According to this disclosure, the accuracy of smoke and flame detection in waste pits can be improved. [Brief explanation of the drawing]

[0011] [Figure 1A] This is a top view showing an example of the configuration of a waste pit and its ancillary facilities. [Figure 1B] This is a side view showing an example of the configuration of a waste pit and its ancillary facilities. [Figure 2] This figure shows an example of the configuration of a detection system. [Figure 3]This figure shows an example of the functional configuration of a detection device included in a detection system. [Figure 4A] This figure shows an example of a detection prompt. [Figure 4B] This is the detection prompt that follows Figure 4A. [Figure 4C] This figure shows an example of a prompt for requesting a water discharge determination. [Figure 5] This figure shows an example of a flame detection result display screen. [Figure 6-1] This is a flowchart illustrating one example of a detection method. [Figure 6-2] This is a flowchart following Figure 6-1. [Figure 6-3] This is a flowchart following Figure 6-2. [Figure 6-4] This is a flowchart following Figure 6-3. [Modes for carrying out the invention]

[0012] Hereinafter, an embodiment of this disclosure will be described with reference to the drawings. For the sake of explanation and notation, the configurations in each drawing have been omitted, simplified, or distorted as necessary.

[0013] <Garbage Pit> Figures 1A and 1B show an example of the configuration of the waste pit 110 and its ancillary facilities. The x, y, and z coordinates shown in Figures 1A and 1B indicate the orientation relationship in each figure. The symbols representing the x, y, and z coordinates are as follows: a black circle inside a white circle corresponds to an arrow line pointing from the back of the page to the front, and a cross mark inside a white circle corresponds to an arrow line pointing from the front of the page to the back.

[0014] Figure 1A is a top view (viewed from the positive z-axis direction toward the negative z-axis direction) showing an example of the configuration of the waste pit 110 and its ancillary facilities. Figure 1B is a side view (viewed from the negative y-axis direction toward the positive y-axis direction) showing an example of the configuration of the waste pit 110 and its ancillary facilities.

[0015] In this embodiment, an example is given where a plane parallel to the horizontal plane (x-y plane) in the garbage pit 110 is divided into a plurality of regions. Note that such division of the garbage pit 110 is for operation management, and there are no boundary lines or partitions at each of the boundaries of these plurality of regions (the two-dot chain line shown in FIG. 1A is not a real line). In FIG. 1A, in the width direction (direction parallel to the x-axis) of the garbage pit 110, the region in the garbage pit 110 is divided into four equal parts, and in the length direction (direction parallel to the y-axis) of the garbage pit 110, the region in the garbage pit 110 is divided into nine equal parts, and an example is given where it is divided into a total of 36 regions. Also, in FIG. 1, it is assumed that A, B, ···, D are assigned in order as identification information of the positions of these plurality of regions in the width direction of the garbage pit 110 from the negative direction to the positive direction side of the x-axis in FIG. 1A. Further, in FIG. 1, it is assumed that 1, 2, ···, 9 are assigned in order as identification information of the positions of these plurality of regions in the length direction of the garbage pit 110 from the negative direction to the positive direction side of the y-axis in FIG. 1A. And in this embodiment, an example is given where information obtained by combining in this order the identification information of the position in the width direction (A, B, ···, D) of the garbage pit 110 and the identification information of the position in the length direction (1, 2, ···, 9) of the garbage pit 110 is the identification information of the plurality of regions. Specifically, for example, the identification information of the region at the lower left corner (the region with the smallest x-axis coordinate and y-axis coordinate respectively) in FIG. 1A is "A1", and the identification information of the region at the upper right corner (the region with the largest x-axis coordinate and y-axis coordinate respectively) is D9. In the following description, the identification information of the plurality of regions that divide a plane parallel to the horizontal plane (x-y plane) in the garbage pit 110 is referred to as an address as needed.

[0016] In determining the address as described above, it is not necessary to divide all regions of a plane parallel to the horizontal plane (x-y plane) within the garbage pit 110 (that is, regions where no address is assigned within the garbage pit 110 may be included). Also, within the region of the garbage pit 110, for example, there may be a receiving area which is a region for receiving waste, and a transfer area which is a region where waste is temporarily placed when moving waste from the receiving area to a stirring area (a region for stirring waste). However, in this embodiment, for simplicity of explanation, a case where all regions within the garbage pit 110 are stirring areas is illustrated. If there are areas other than the stirring area within the garbage pit 110, the above-described address may be assigned only to the stirring area, or the address may be assigned to both the stirring area and other areas. Also, generally, a partition wall is installed at the boundary between the receiving area and another area adjacent to the receiving area.

[0017] Also, in FIGS. 1A and 1B, a case where two cranes 120a to 120b are installed in the garbage pit 110 is illustrated. However, the number of cranes installed for the garbage pit 110 is not limited to 2, and may be 1 or 3 or more. In this embodiment, a case where the two cranes 120a to 120b are cranes with the same configuration is illustrated. However, the two cranes 120a to 120b may be cranes with different configurations.

[0018] Figure 1A illustrates a case where cranes 120a to 120b are equipped with traveling devices 121a to 121b and traversing devices 122a to 122b. In Figure 1A, the traversing direction of cranes 120a to 120b is the width direction of the waste pit 110 (parallel to the x-axis), and the traveling direction of cranes 120a to 120b is the length direction of the waste pit 110 (parallel to the y-axis). In the following explanation, the traversing direction of cranes 120a to 120b will be abbreviated as "traversing direction" as needed, and the traveling direction of cranes 120a to 120b will be abbreviated as "traveling direction" as needed. Note that in Figure 1B, for the sake of notation, only the bottom of traveling devices 121a to 121b is shown. Also in Figure 1B, traveling device 121b, which is hidden by traveling device 121a, is shown with a dashed line.

[0019] Figure 1A illustrates a case where the running rail 130 extends on both sides of the width direction of the waste pit 110 at a position higher than the upper end of the waste pit 110 (the end on the positive z-axis side). The direction in which the running rail 130 extends is parallel to the direction of travel. The running devices 121a to 121b move (travel) on the running rail 130. The running devices 121a to 121b can be implemented, for example, by various known girders.

[0020] Furthermore, Figure 1A illustrates the case where the traverse rails 123a to 123b extend within the traveling devices 121a to 121b. The direction in which the traverse rails 123a to 123b extend is parallel to the traverse direction. The traverse devices 122a and 122b move (traverse) on the traverse rails 123a and 123b, respectively. The traverse devices 122a to 122b are equipped with buckets 124a and 124b. As shown in Figure 1B, the buckets 124a and 124b grasp and release (open and close) the waste stored in the waste pit 110. The waste is transported by the movement of the cranes 120a and 120b (traveling devices 121a and 122b, and traverse devices 122a and 122b) while the buckets are grasping the waste. The traverse devices 122a to 122b can be implemented, for example, by various known club trolleys. The cranes 120a and 120b themselves are not limited to cranes having the aforementioned configuration, but are implemented using known cranes that transport waste stored in the waste pit 100 to an incinerator or the like.

[0021] Furthermore, in Figures 1A and 1B, this embodiment illustrates a case where two imaging devices 140a to 140b are installed in the waste pit 110. However, the number of imaging devices installed in the waste pit 110 is not limited to two; it may be one or three or more. In this embodiment, the case where the two imaging devices 140a to 140b have the same configuration is illustrated. However, the two imaging devices 140a to 140b may have different configurations. The imaging devices 140a to 140b may be installed, for example, in the crane operation room (where the operator operates the cranes 120a and 120b), but they may also be installed in a different location from the crane operation room. The crane operation room may be installed outside the cranes 120a and 120b, or it may be installed on the cranes 120a and 120b (for example, on the traveling device). The imaging devices 140a and 140b may be, for example, ITV (Industrial television: surveillance cameras). In Figure 1B, imaging device 140b, which is hidden by imaging device 140a, is shown with a dashed line.

[0022] The imaging devices 140a and 140b capture images that include the waste 190 stored in the waste pit 110. The waste 190 stored in the waste pit 110 includes the waste 190 piled up in the waste pit 110. The imaging devices 140a and 140b may also capture images that include the waste being grasped by the buckets 124a and 124b of the cranes 120a and 120b, as images that include the waste 190 stored in the waste pit 110. In this embodiment, the case in which the images captured by the imaging devices 140a and 140b are moving images is illustrated. However, the images captured by the imaging devices 140a and 140b may also be still images. If the images captured by the imaging devices 140a and 140b are still images, the imaging devices 140a and 140b may capture still images at a timing specified from the outside, or at a preset timing (for example, at a certain period). For example, the image input to the language model 221, described later, may be a still image. Furthermore, the still image may be a frame image extracted from a moving image. Also, the images captured by the imaging devices 140a to 140b may be color images or grayscale images. In this embodiment, we illustrate the case where imaging devices 140a and 140b capture images based on visible light.

[0023] The shooting conditions (angle of view, etc.) of the shooting devices 140a to 140b may be determined so that images of all areas within the waste pit 110 (images of all waste 190 accumulated in an exposed state within the waste pit 110) can be obtained. However, the shooting area of ​​the shooting devices 140a to 140b does not have to include areas of some areas within the waste pit 110. For example, if there are areas where the generation of at least one of smoke and flame does not need to be considered, images of those areas do not need to be taken. Also, images of areas that are in the blind spots of the shooting devices 140a and 140b due to the position of the cranes 120a and 120b will not be taken by the shooting devices 140a and 140b until the cranes 120a and 120b are moved. Note that the installation positions of the shooting devices 140a to 140b are not limited to the positions exemplified in Figures 1A and 1B, as long as the desired area within the waste pit 110 can be photographed. In the following explanation, images captured by imaging devices 140a to 140b will be referred to as pit images as needed, and image data representing pit images will be referred to as pit image data as needed.

[0024] Furthermore, in Figures 1A and 1B, this embodiment illustrates a case where two water discharge devices 150a to 150b are installed in the waste pit 110. However, the number of water discharge devices installed in the waste pit 110 is not limited to two; it may be one or three or more. In this embodiment, the case where the two water discharge devices 150a to 150b have the same configuration is illustrated. However, the two water discharge devices 150a to 150b may have different configurations. Note that in Figure 1B, water discharge device 150b, which is hidden by water discharge device 150a, is shown with a dashed line.

[0025] The water discharge devices 150a to 150b are devices that discharge water onto the waste (and the smoke and flames generated from it) stored in the waste pit 110. The waste stored in the waste pit 110 (waste to be discharged with water) includes the waste 190 piled up in the waste pit 110. Waste being held by the cranes 120a and 120b (buckets 124a and 124b), and waste falling from the cranes 120a and 120b (buckets 124a and 124b) may also be included in the waste stored in the waste pit 110 (waste to be discharged with water). The water discharge devices 150a to 150b have, for example, water cannons. In this case, for example, water pumped up by pumps 151a to 151b is supplied to the water cannons via solenoid valves 152a to 152b, etc., and water is sprayed from the water cannons. The water discharge devices 150a to 150b may be able to change the direction of water discharge (direction of water injection). Furthermore, the range in which water can be discharged by the water discharge devices 150a to 150b may be defined so that images of all addresses within the waste pit 110 (images of all waste 190 stored in an exposed state within the waste pit 110) can be obtained. However, the range in which water can be discharged by the water discharge devices 150a to 150b does not necessarily include the areas of some addresses within the waste pit 110. For example, if there are addresses where the generation of at least one of smoke and flame does not need to be considered, those addresses may not be included in the range in which water can be discharged by the water discharge devices 150a to 150b. Also, the water discharge devices 150a to 150b may be operated based on worker operation or automatically controlled. This embodiment illustrates a case where both are possible. Note that the water discharge devices 150a to 150b themselves may be existing equipment installed as ancillary facilities to the waste pit 110. In this embodiment, the example shows a case where the amount of water discharged from the water discharge devices 150a to 150b is fixed, but the amount of water discharged from the water discharge devices 150a to 150b may also be variable. Furthermore, the installation positions of the water discharge devices 150a to 150b are not limited to the positions exemplified in Figures 1A and 1B, as long as water can be discharged to a desired area within the waste pit 110.

[0026] Furthermore, Figures 1A and 1B illustrate an example in this embodiment where one ultraviolet sensor 160 is installed in the waste pit 110. However, the number of ultraviolet sensors installed in the waste pit 110 is not limited to one, and may be two or more.

[0027] The ultraviolet sensor 160 is a sensor that detects ultraviolet light emitted from waste 190 stored in the waste pit 110. The waste 190 stored in the waste pit 110 includes waste 190 piled up in the waste pit 110. In this embodiment, an example is given in which the ultraviolet sensor 160 has an ultraviolet camera that takes an ultraviolet image. The ultraviolet image is an image having pixel values ​​corresponding to the intensity of ultraviolet light. However, the ultraviolet sensor 160 does not have to be a sensor that takes an ultraviolet image. For example, the ultraviolet sensor may be a sensor that only detects the presence or absence of ultraviolet light (however, in this case, the location of the waste that is the source of the ultraviolet light cannot be determined from the detection result of the ultraviolet sensor alone).

[0028] The range in which the ultraviolet sensor 160 can detect ultraviolet light may be defined so that the waste 190 in all addresses within the waste pit 110 is included in the detection area. However, the range in which the ultraviolet sensor 160 can detect ultraviolet light does not have to include the areas of some addresses within the waste pit 110. For example, if there is an address where the generation of at least one of smoke and flame does not need to be considered, that address does not have to be included in the range in which the ultraviolet sensor 160 can detect ultraviolet light. Also, Figures 1A and 1B illustrate the case where the ultraviolet sensor 160 is installed on the ceiling of the waste pit 110. However, the ultraviolet sensor 160 does not have to be installed on the ceiling of the waste pit 110.

[0029] Furthermore, in Figures 1A and 1B, this embodiment illustrates a case where one infrared sensor 170 is installed in the waste pit 110. However, the number of infrared sensors installed in the waste pit 110 is not limited to one, and may be two or more.

[0030] The infrared sensor 170 is a sensor that detects infrared radiation emitted from waste 190 stored in the waste pit 110. The waste 190 stored in the waste pit 110 includes waste 190 piled up in the waste pit 110. In this embodiment, an example is given in which the infrared sensor 170 has an infrared camera that takes infrared images. The infrared image is an image having pixel values ​​corresponding to the intensity of infrared radiation. However, the infrared sensor 170 does not have to be a sensor that takes infrared images. For example, the infrared sensor may be a sensor that only detects the presence or absence of infrared radiation (however, in this case, the location of the waste that is the source of the infrared radiation cannot be determined from the detection result of the infrared sensor alone).

[0031] The range in which the infrared sensor 170 can detect infrared radiation may be defined so that the waste 190 in all addresses within the waste pit 110 is included in the detection area. However, the range in which the infrared sensor 170 can detect infrared radiation does not have to include the areas of some addresses within the waste pit 110. For example, if there is an address where the generation of at least one of smoke and flame does not need to be considered, that address does not have to be included in the range in which the infrared sensor 170 can detect infrared radiation. Also, Figures 1A and 1B illustrate the case where the infrared sensor 170 is installed on the ceiling of the waste pit 110. However, the infrared sensor 170 does not have to be installed on the ceiling of the waste pit 110.

[0032] The ultraviolet sensor 160 and infrared sensor 170 themselves may be known types. In this embodiment, we illustrate a case where the ultraviolet sensor 160 and infrared sensor 170 are examples of invisible light sensors (sensors that detect invisible light). However, invisible light sensors are not limited to ultraviolet sensors and infrared sensors. For example, an invisible light sensor may be a sensor configured by combining an ultraviolet sensor and an infrared sensor into a single sensor. This sensor outputs a final detection result using, for example, both the detection result of ultraviolet light and the detection result of infrared light. Also, one of the ultraviolet sensor 160 and the infrared sensor 170 does not have to be installed in the waste pit 110.

[0033] In Figure 1B, when the waste input gate (input door) 180 is opened, waste can be discharged into the waste pit 110 from a packer truck or the like (through the opening of the waste input gate 180). As shown in Figure 1B, the imaging devices 140a to 140b may be installed above the waste input gate 180. This is preferable because it allows for more reliable imaging of the characteristics of the waste being input into the waste pit 110 and the characteristics of the waste accumulated in the waste pit 110. The water discharge devices 150a to 150b may also be installed above the waste input gates 160a to 160e. This is preferable because it allows for easy and reliable water discharge onto the waste accumulated in the waste pit 110, regardless of the amount of waste (height of the waste layer) accumulated in the waste pit 110.

[0034] <Overall configuration of the detection system> Figure 2 shows an example of the configuration of the detection system of this embodiment. In Figure 2, this embodiment illustrates a case where the detection system includes a detection device 210.

[0035] In Figure 2, the detection device 210, the language model server 220, and the external control device 230 are connected to each other so as to be able to communicate with one another via a network 250, for example, the Internet. At least one of the detection device 210, the language model server 220, and the external control device 230 may reside on the cloud. The detection device 210 and at least one of the language model server 220 and the external control device 230 may be connected to each other so as to be able to communicate with one another via a network other than the Internet, or they may communicate via wired communication or wireless communication. The language model server 220 may also be included in the detection device 210. In Figure 2, the detection device 210 and the processing device 240 are connected to each other so as to be able to communicate with one another via a network 260, for example, the LAN (Local Area Network). The detection device 210 and the processing device 240 may be connected to each other so as to be able to communicate with one another via a network other than the LAN, or they may be connected so as to be able to communicate without a network, or they may communicate via wired communication or wireless communication.

[0036] In this embodiment, we illustrate a case where the detection device 210 is a device included in a computer system installed in the central control room of a waste treatment facility. Furthermore, we illustrate a case where the processing device 240 is included in a computer system installed in the crane operation room of a waste treatment facility. Also, we illustrate a case where the external control device 230 is installed in a location different from the area where the waste pit 110 is located. Specifically, we illustrate a case where the external control device 230 is a device for remotely monitoring the waste treatment facility and is installed outside the premises of the waste treatment facility having the waste pit 110. Note that the functions of the detection device 210, as described later, may be included in a computer system installed in the crane operation room (e.g., the processing device 240) in addition to, or instead of, the computer system installed in the central control room of the waste treatment facility, or in a computer system installed in a facility for remotely monitoring the waste treatment facility (e.g., the external control device 230). If the functions of the detection device 210, as described later, are not included in the computer system installed in the central control room of the waste treatment facility, then the device installed in the central control room becomes an external device from the perspective of the device having the functions of the detection device 210, as described later.

[0037] The language model server 220, in response to requests from external devices such as the detection device 210, causes a language model trained by machine learning using a large amount of data to perform natural language processing and transmits the results of the processing to the external device. By using the language model, for example, machine translation, text generation, and answering questions can be performed automatically. The language model can be any model that can recognize images. For example, the language model may be a large language model (LLM) such as GPT4-o(registered trademark) or LLaVA 1.5(registered trademark). The language model may be trained with a dataset using prompts, or it may be trained with a dataset through fine tuning.

[0038] The detection device 210 includes, for example, one or more hardware processors such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit), and one or more memory modules such as RAM (Random Access Memory) and ROM (Read Only Memory). It performs various calculations by executing one or more programs stored in the memory using the one or more hardware processors. The detection device 210 also includes, for example, a NIC (Network Interface Card) for connecting the detection device 210 to a network.

[0039] Furthermore, a storage medium that can be read by the detection device 210 (computer) may be connected to the detection device 210. Note that this storage medium is not limited to a storage medium that can only read data; it may also be a storage medium that can be read and written to by the detection device 210 (computer). Additionally, an output device such as a computer display may be connected to the detection device 210. Furthermore, an input device such as a keyboard or mouse may be connected to the detection device 210. Information input to the detection device 210 may also be performed using a GUI (Graphical User Interface).

[0040] Furthermore, the detection device 210 may be implemented using dedicated hardware such as an ASIC (Application Specific Integrated Circuit) and an FPGA (Field Programmable Gate Array). The detection device 210 may also include a PLC (Programmable Logic Controller). The language model server 220, external control device 230, and processing device 240 can also be implemented using hardware similar to that of the detection device 210. Depending on the performance of the detection device 210, the language model may be stored in the detection device 210. In this case, the language model server 220 may not be necessary. Furthermore, while this embodiment illustrates the case where a language model is used, it is not necessarily required to use a language model (this point will be discussed later). In this case as well, the language model server 220 may not be necessary.

[0041] <Functional Configuration of the Detection System> Figure 3 shows an example of the functional configuration of a detection device 210 included in the detection system. In Figure 3, this embodiment illustrates a case where the detection device 210 comprises an acquisition unit 211, a first detection unit 212, a second detection unit 213, a control unit 214, a storage unit 215, a communication unit 216, and an input receiving unit 217.

[0042] <<Acquisition part 211>> The acquisition unit 211 acquires estimation image data showing images of the waste pit 110 where waste is stored. In this embodiment, an example is given where the acquisition unit 211 acquires images of the inside of the pit taken by the imaging devices 140a to 140b. Here, the estimation image data is image data showing images of the waste pit 110 used to estimate the generation of at least one of smoke and flames in the waste pit 110. In the following description, the images of the inside of the pit used to estimate the generation of at least one of smoke and flames in the waste pit 110 will be referred to as estimation images of the inside of the pit, as necessary, and the image data of the inside of the pit showing estimation images of the inside of the pit will be referred to as estimation images of the inside of the pit, as necessary. Also, in the following description, at least one of smoke and flames in the waste pit 110 will be referred to as smoke and flames in the waste pit 110, or simply smoke and flames, as necessary.

[0043] The acquisition unit 211 acquires the data by receiving, for example, pit-in-estimation image data transmitted from a communication device built into the imaging devices 140a to 140b, or from a communication device connected to the imaging devices 140a to 140b. Communication between the detection device 210 and the communication device may be via wired communication, wireless communication, or network communication.

[0044] Furthermore, in this embodiment, we illustrate a case in which the acquisition unit 211 acquires ultraviolet sensor data indicating the detection result of ultraviolet light in the waste pit 110, which has been detected by the ultraviolet sensor 160. As mentioned above in the <waste pit> section, in this embodiment, we illustrate a case in which the ultraviolet sensor 160 has an ultraviolet camera that takes ultraviolet images. In this case, the ultraviolet sensor data includes ultraviolet image data showing an ultraviolet image of the waste pit 110.

[0045] Furthermore, in this embodiment, we illustrate a case in which the acquisition unit 211 acquires infrared sensor data indicating the detection result of infrared radiation inside the waste pit 110, which has been detected by the infrared sensor 170. As mentioned above in the <Waste Pit> section, in this embodiment, we illustrate a case in which the infrared sensor 170 has an infrared camera that takes infrared images. In this case, the infrared sensor data includes infrared image data showing an infrared image of the waste pit 110.

[0046] Furthermore, this embodiment illustrates a case in which the acquisition unit 211 acquires a detection prompt. The detection prompt is an example of a prompt that causes the language model 221 to output information corresponding to the smoke / flame detection result in the waste pit 110. In this embodiment, an example is given of a case in which the detection prompt is a prompt that explicitly instructs the language model 221 to output the smoke / flame detection result. That is, in this embodiment, an example is given of a prompt that causes the language model 221 to output information corresponding to the smoke / flame detection result, which is a prompt that causes the language model 221 to output information indicating the smoke / flame detection result. However, the detection prompt is not limited to such a prompt. For example, the detection prompt may include an instruction that causes the language model 221 to output information that requires the language model 221 to detect smoke / flame without explicitly instructing the detection of smoke / flame. For example, the detection prompt may be a prompt that instructs the language model 221 to notify the occurrence of smoke / flame with a buzzer or the like. In this case, the presence or absence of notification by a buzzer or the like is an example of information corresponding to the smoke / flame detection result. In this embodiment, an example is given where the detection device 210 (control unit 214) determines whether or not to issue a notification (see the sections on <<<First Control Process>>> and <<<Second Control Process>>> described later). The detection prompt may be input to the language model 221 as a user prompt or as a system prompt. In the latter case, the detection prompt may be initially set as the system prompt of the language model 221.

[0047] Furthermore, when estimation pit image data is input to the language model 221, if the language model 221 outputs information corresponding to the smoke / flame detection result in the waste pit 110, it is not necessarily required to use a prompt. However, in this embodiment, an example is given of using a detection prompt to more reliably enable the language model 221 to perform this action. In addition, in this embodiment, an example is given of using a detection prompt to enable the language model 221 to estimate a wider range of information based on the estimation pit image data.

[0048] Figures 4A and 4B show an example of a detection prompt 400. In Figures 4A and 4B, this embodiment illustrates a case where the detection prompt 400 includes a first prompt 401 that instructs the language model 221 to estimate the type (category) of waste stored in the waste pit 110, in addition to the presence or absence of smoke and flames, based on the image data inside the estimation pit. Furthermore, this example illustrates a case where the first prompt 401 also includes an instruction to calculate the confidence level of each estimation result.

[0049] Alternatively, the language model 221 may be trained using a dataset in which pit image data and correct labels are correlated. The correct labels include, for example, information indicating whether smoke or flames are being generated, and information indicating whether what is being generated is smoke or flames. The correct labels also include, for example, information indicating the type of waste.

[0050] Furthermore, in this embodiment, we illustrate a case in which the detection prompt 400 includes a second prompt 402 that instructs the system to estimate the presence or absence of smoke and flames and the type of waste for each address of the waste pit 110, and a third prompt 403 that describes the address.

[0051] Furthermore, in this embodiment, we illustrate a case where the detection prompt 400 includes a fourth prompt 404 that instructs the language model 221 to estimate whether smoke or flame is being generated. We also illustrate a case where the fourth prompt 404 includes an instruction to calculate the confidence level of the estimation result.

[0052] Furthermore, in this embodiment, we illustrate a case where the fifth prompt 405, which instructs the language model 221 on the format in which it will output the estimation result, is included in the detection prompt 400.

[0053] Note that the prompts included in the detection prompt 400 are not limited to the prompts exemplified in Figure 4A. For example, the detection prompt 400 may include prompts that instruct the language model 221 to perform various control operations. In this case, for example, by using function calling, the language model 221 may call a control function and output it to an external device, and the external device may perform control based on that function. In this case, the control may include the control of the operation of cranes 120a to 120b. This control may be, for example, a control that reduces the bias of the types of waste accumulated in the waste pit 110 by address after the operation of cranes 120a to 120b is controlled compared to before the operation was controlled. By controlling the operation of cranes 120a to 120b in this way, agitation such as scattering agitation or moving agitation may be performed. The external device may also be a server (for example, an MCP (Model Context Protocol) server) that can be accessed via a network (for example, network 250). In this case, the language model 221 may make a function call to the server (the external device), and control of the external control device (for example, control of the operation of cranes 120a to 120b) may be performed via the server.

[0054] Furthermore, in this embodiment, we illustrate the case in which the acquisition unit 211 acquires a prompt for requesting a water discharge determination. The prompt for requesting a water discharge determination is an example of a prompt that causes a language model 221, which is an example of a determination unit, to determine whether or not to control the water discharge devices 150a to 150b, which are examples of water discharge units, to discharge water at the smoke and flames, based on the smoke and flame detection results in the garbage pit 110, which are performed by the first detection unit 212 and the second detection unit 213 as described later. Note that the determination unit is not limited to the language model 221. For example, the determination unit may perform the determination using a rule-based method. The determination unit may be located inside the detection device 210 or outside the detection device 210. Alternatively, the control unit 214, which will be described later, may determine whether or not to control the water discharge devices 150a to 150b itself. In this case, the determination unit is not necessary. In the following description, the detection results of smoke and flames in the waste pit 110, performed by the first detection unit 212 and the second detection unit 213, will be referred to as the detection results of the first detection unit 212 and the detection results of the second detection unit 213, respectively, as needed. When the first detection result information indicating the detection result of the first detection unit 212 or the second detection result information indicating the detection result of the second detection unit 213 is input to the language model 221, it is not necessarily required to use a prompt if the language model 221 outputs at least whether or not water discharge is necessary. However, in this embodiment, an example is given of using a water discharge decision request prompt to make the language model 221 perform this more reliably. In addition, in this embodiment, an example is given of using a water discharge decision request prompt to cause the language model 221 to estimate more information based on the first detection result information and the second detection result information. The water discharge decision request prompt may be input to the language model 221 as a user prompt or as a system prompt. In the latter case, the prompt for requesting a water discharge determination may be initially set as the system prompt for language model 221.

[0055] Figure 4C shows an example of a prompt 410 for requesting a water discharge determination. In Figure 4C, this embodiment illustrates a case where the first prompt 411, which instructs the language model 221 to determine whether or not to discharge water onto the garbage pit 110 based on information indicating the smoke and flame detection results in the garbage pit 110 (first detection result information and second detection result information in this embodiment), is included in the prompt 410 for requesting a water discharge decision. Furthermore, this example illustrates a case where the first prompt 411 also includes an instruction to determine the address to which water should be discharged.

[0056] Furthermore, in this embodiment, we illustrate a case in which the prompt 410 for requesting a water discharge determination includes a second prompt 412 explaining that the smoke and flame detection results in the waste pit 110 (first detection result information and second detection result information in this embodiment) show results for each address, and a third prompt 413 explaining the address. Furthermore, in this embodiment, a fourth prompt 414 is provided in the prompt 410 for requesting a water discharge determination, explaining that the smoke and flame detection results in the waste pit 110 (first detection result information and second detection result information in this embodiment) include results based on a trained model and results based on an invisible light sensor, and that images of the waste pit at the time of determination (images taken inside the pit) are also provided.

[0057] Furthermore, in this embodiment, we illustrate a case where the fifth prompt 415, which instructs the language model 221 on the format in which the judgment result should be output, is included in the water discharge judgment request prompt 410. Note that the detection prompt 400 and the water discharge determination request prompt 410 are intended to make the explanation easier to understand, and it is not necessary to provide the language model with the content shown in Figures 4A to 4C as is. They may be modified or changed as appropriate to be applicable to actual operation.

[0058] The operator of the detection device 210 may input information indicating the contents of the detection prompt 400 and the water discharge decision request prompt 410 to the detection device 210 by operating an input device connected to the detection device 210. In this case, the acquisition unit 211 acquires the information indicating the contents of the detection prompt 400 and the water discharge decision request prompt 410 that has been input in this manner. Alternatively, the information indicating the contents of the detection prompt 400 and the water discharge decision request prompt 410 may be stored on an external storage medium of the detection device 210. In this case, the acquisition unit 211 may acquire the information indicating the contents of the detection prompt 400 and the water discharge decision request prompt 410 by reading it from the external storage medium of the detection device 210. Alternatively, the acquisition unit 211 may acquire the information indicating the contents of the detection prompt 400 and the water discharge decision request prompt 410 by receiving it from an external device.

[0059] Furthermore, in this embodiment, we illustrate a case in which the acquisition unit 211 acquires various types of information that the detection device 210 needs to acquire from the outside in order to process, in addition to the information described above. For example, the acquisition unit 211 may acquire crane position information (information indicating the current position of cranes 120a and 120b (buckets 124a and 124b)) which will be described later in the first control processing section of the control unit 214.

[0060] <<First detection unit 212>> In Figure 3, the first detection unit 212 receives captured image data showing the captured image and detects the occurrence of smoke and flames in the waste pit 110 based on a trained model that estimates the occurrence of smoke and flames in the waste pit 110 and the estimated captured image data acquired by the acquisition unit 211.

[0061] In this embodiment, an example is given where the trained model is a trained model that takes pit image data as input and estimates the occurrence of smoke and flames in the garbage pit 110. The number of trained models used by the first detection unit 212 may be one or two or more. In this embodiment, an example is given where the trained models used by the first detection unit 212 include a language model 221 and a first trained model. The first trained model is a trained model that has learned the relationship between captured image data as training data, which shows captured images in the garbage pit 110, and training data that indicates whether or not smoke and flames are included in the subject of the captured image shown by the captured image data.

[0062] In this embodiment, the first trained model is a trained model that has learned the relationship between pit image data as training data and training data indicating whether or not smoke or flames are included in the subject of the captured image shown by the pit image data through supervised learning, and is an example of a trained model created for estimating the occurrence of smoke or flames in the garbage pit 110. In this embodiment, the training data further includes information indicating whether the subject is smoke or flames, and the position of the subject in the captured image (e.g., pixel number).

[0063] The first trained model may be, for example, an object detection model such as YOLO (You Only Look Once) or CNN (Convolutional Neural Networks). In this embodiment, we will illustrate the case where the first trained model is a trained model that also outputs the confidence level of the detection result. In this embodiment, for the sake of simplicity, we will illustrate the case where there is one first trained model (for example, YOLO), but there may be two or more first trained models (for example, both YOLO and CNN). The two or more first trained models may be the same type of trained model or different types of trained models. Since machine learning itself, such as supervised learning, is implemented using known techniques, we will omit a detailed explanation of the training method for the first trained model here.

[0064] Furthermore, this embodiment illustrates a case where the first trained model detects the location of smoke and flames based on the image inside the pit. Furthermore, this embodiment illustrates a case where the language model 221 detects the location of smoke and flames based on the detection prompt 400 and the image inside the pit. In this embodiment, it is illustrated that the location of smoke and flames is represented by an address. Furthermore, this embodiment illustrates a case where the detection result of smoke and flames in the first trained model and the language model 221 includes a confidence level of the detection result.

[0065] The first detection unit 212 may, for example, adopt the detection result output from the trained model as correct if the confidence level output from the trained model exceeds a predetermined threshold. In this case, if the confidence level output from the trained model is below the predetermined threshold, the first detection unit 212 may discard the detection result for which the confidence level is below the predetermined threshold. However, this is not always necessary. For example, the first detection unit 212 may adopt the detection result output from the trained model as correct regardless of the confidence level.

[0066] Furthermore, the first detection unit 212 may unconditionally detect the occurrence of smoke and flames in the waste pit 110 based on all trained models (language model 221 and first trained model) that the first detection unit 212 can use. However, doing so may increase the processing load on the detection device 210, potentially preventing rapid detection. Also, from the viewpoint of suppressing an increase in detection costs, it may be preferable to suppress the frequent use of the language model 221. In this case, the frequency at which the estimated pit image data is input to the language model 221 may be lower than the frequency at which the estimated pit image data is input to the first trained model. For example, the estimated pit image data may be input to the first trained model each time the acquisition unit 211 acquires the estimated pit image data.

[0067] From these illustrative viewpoints, this embodiment illustrates a case in which the first detection unit 212 detects the generation of smoke and flames in the waste pit 110 when any of the following conditions 1 to 5 are met.

[0068] <<<Condition 1>> The first condition is met when the first trained model estimates the occurrence of smoke and flames in the waste pit 110 (the first trained model detects the occurrence of smoke and flames), or when the second detection unit 213, described later, detects the occurrence of smoke and flames in the waste pit 110. In this case, the first detection unit 212 causes the language model 221 to estimate the occurrence of smoke and flames. In this embodiment, an example is given in which the first detection unit 212 outputs estimation pit image data to the input receiving unit 217 when the first condition is met. In this case, the first detection unit 212 may output estimation pit image data to the input receiving unit 217 showing one estimation pit image taken at the closest time to when it was determined that the first condition was met. If the estimation pit image data is a moving image, the first detection unit 212 may output estimation pit image data showing one frame image to the input receiving unit 217. Furthermore, the same applies when the following conditions 2 to 5 are met: the estimation pit image data showing a single estimation pit image may be output to the input receiving unit 217 in this manner.

[0069] <<<Second condition>>> The second condition is met when the reliability of the estimation of smoke and flame generation in the waste pit 110 by the first trained model is above a predetermined value and below a predetermined value. In this embodiment, when the second condition is met, the first detection unit 212 outputs the estimation pit image data to the input reception unit 217, which will be described later, as an example.

[0070] In this way, for example, if the confidence level of the first trained model's estimation of smoke and flame generation is neither high nor low (i.e., if the first trained model is not confident in its inference), the language model 221 can be made to estimate the generation of smoke and flame. In this embodiment, even in this case, we illustrate the case in which the first detection unit 212 outputs the estimation pit image data to the input reception unit 217, which will be described later.

[0071] Furthermore, the first detection unit 212 may, without setting a lower limit, cause the language model 221 to estimate the occurrence of smoke and flames in the waste pit 110 when the reliability of the estimation of smoke and flame occurrence in the waste pit 110 by the first trained model is below a predetermined value. However, in order to reduce the frequency of use of the language model 221, in this embodiment, as described above, an example is given in which the language model 221 is caused to estimate the occurrence of smoke and flames when the reliability of the estimation of smoke and flame occurrence in the waste pit 110 by the first trained model is above a predetermined value and below a predetermined value.

[0072] <<<Third Condition>>> The third condition is met when, after the water discharge unit discharges water onto the smoke and flames, smoke and flames are detected based on the first trained model, or when the second detection unit 213, described later, detects smoke and flames. In this embodiment, when the third condition is met, the first detection unit 212 outputs the estimation pit image data to the input receiving unit 217, described later, as an example.

[0073] <<<Condition 4>>> The fourth condition is met when the control unit 214, described later, issues an instruction to detect smoke or flames in the waste pit 110. In the following description, this instruction will be referred to as a detection instruction as needed. In this case, the first detection unit 212 performs smoke or flame detection using a type of trained model that is specified to be used based on the detection instruction. That is, in this case, the first trained model may be used. In this embodiment, when the type of trained model that is specified to be used based on the detection instruction is the language model 221, we will illustrate the case in which the first detection unit 212 outputs estimation pit image data to the input receiving unit 217, described later.

[0074] <<<Condition 5>>> The fifth condition is when a predetermined periodic timing arrives. In this embodiment, when the fifth condition is met, the first detection unit 212 detects the generation of smoke and flames in the waste pit 110 based on the first trained model, and does not cause the language model 221 to estimate the generation of smoke and flames.

[0075] <<Second detection unit 213>> In Figure 3, the second detection unit 213 uses an invisible light sensor to detect the generation of smoke and flames in the waste pit 110. As mentioned above in the section on <waste pit>, in this embodiment, the ultraviolet sensor 160 and the infrared sensor 170 are invisible light sensors, and the example given is that the ultraviolet sensor 160 and the infrared sensor 170 have an ultraviolet camera and an infrared camera, respectively.

[0076] The second detection unit 213 detects the occurrence of smoke and flames in the waste pit 110 based on the signal output from the ultraviolet sensor 160 (ultraviolet image data in this embodiment). The method for detecting the occurrence of smoke and flames in the waste pit 110 is realized by using a known method for detecting the occurrence of smoke and flames using an ultraviolet sensor, so a detailed explanation is omitted here. In this embodiment, an example is given in which the second detection unit 213 detects the location of the smoke and flames based on an ultraviolet image of the waste pit 110. For example, the second detection unit 213 detects the location of the smoke and flames based on the pixel position that is the basis for the occurrence of smoke and flames. For example, the relationship between each pixel of the ultraviolet image and its position in real space in the waste pit 110 may be investigated in advance, and information indicating the investigated relationship may be stored in the storage unit 215. In this case, the second detection unit 213 may use this information to detect the location of the smoke and flames. In this embodiment, an example is given in which the location of the smoke and flames is represented by an address.

[0077] Furthermore, the second detection unit 213 detects the occurrence of smoke and flames in the garbage pit 110 based on a signal output from the infrared sensor 170 (in this embodiment, infrared image data). The method for detecting the occurrence of smoke and flames in the garbage pit 110 is realized by using a known method for detecting the occurrence of smoke and flames using an infrared sensor, so a detailed explanation is omitted here. In this embodiment, an example is given in which the second detection unit 213 detects the location of the smoke and flames based on an infrared image of the garbage pit 110. For example, the second detection unit 213 detects the location of the smoke and flames based on the pixel position that is the basis for the occurrence of smoke and flames. For example, the relationship between each pixel of the infrared image and its position in real space in the garbage pit 110 may be investigated in advance, and information indicating the investigated relationship may be stored in the storage unit 215. In this case, the second detection unit 213 may use this information to detect the location of the smoke and flames. In this embodiment, an example is given in which the location of the smoke and flames is represented by an address.

[0078] Furthermore, the second detection unit 213 may, similar to the first detection unit 212, further use the language model 221 to detect smoke and flames in the waste pit 110. In this case, the explanation would be, for example, in the explanation of the section "<<First detection unit 212>>", with the first detection unit 212 and the pit image replaced by the second detection unit 213 and either an ultraviolet image or an infrared image, respectively. In this case, a detection prompt 400 may also be used.

[0079] In this embodiment, an example is given in which the second detection unit 213 detects the generation of smoke and flames in the waste pit 110 using the ultraviolet sensor 160 and the infrared sensor 170 when the following sixth and seventh conditions are met. The second detection unit 213 may also detect the generation of smoke and flames in the waste pit 110 using only one of the ultraviolet sensor 160 and the infrared sensor 170.

[0080] <<<Condition 6>>> The sixth condition is met when there is a detection instruction from the control unit 214, which will be described later.

[0081] <<<Condition 7>>> The seventh condition is the arrival of a predetermined period. This period may be a pre-set period. The period during which the first detection unit 212 detects the presence or absence of smoke / flames using the first trained model and the period during which the second detection unit 213 detects the presence or absence of smoke / flames using the ultraviolet sensor 160 and the infrared sensor 170 may be the same or different. Furthermore, the timing at which the first detection unit 212 detects the presence or absence of smoke / flames using the first trained model and the timing at which the second detection unit 213 detects the presence or absence of smoke / flames using the ultraviolet sensor 160 and the infrared sensor 170 may be synchronized.

[0082] <<Control Unit 214>> The control unit 214 executes various control processes in the detection device 210. In this embodiment, an example is given in which the control unit 214 performs control processes including the first to tenth control processes.

[0083] <<<First control process>>> First, an example of the first control process will be explained. The first control process includes control processing related to notifying the results of smoke and flame detection in the waste pit 110.

[0084] When at least one of the first detection unit 212 and the second detection unit 213 detects the generation of smoke or flames in the waste pit 110, the control unit 214 causes the first notification unit to notify the first notification unit of detection result information indicating the respective detection results of the first detection unit 212 and the second detection unit 213. The first notification unit includes, for example, at least one of a display unit (e.g., a computer display), a light-emitting unit (e.g., a lamp), and a sound-emitting unit (e.g., a buzzer). The first notification unit may be, for example, a notification unit that is communicably connected to the detection device 210, a notification unit that is communicably connected to the external control device 230, or a notification unit that is communicably connected to the processing device 240. In this embodiment, when notification is made to a notification unit that is communicatively connected to an external control device 230 and to a notification unit that is communicatively connected to a processing device 240, the detection result information is transmitted to the external control device 230 and the processing device 240 via the communication unit 216, and notification is made in the notification unit (first notification unit) that is communicatively connected to the external control device 230 and the processing device 240.

[0085] If the first notification unit is a display unit, the following processing may be performed as the first control processing. First, the control unit 214 may display on the display unit, in a comparable manner, first area information indicating the first area in the waste pit 101 where the first detection unit 212 detected smoke and flames, and second area information indicating the second area in the waste pit 101 where the second detection unit 213 detected smoke and flames.

[0086] Furthermore, the control unit 214 may display on the display unit in a discriminable manner the area in the waste pit 101 where the first area where the first detection unit 212 detected smoke / flames and the second area where the second detection unit 213 detected smoke / flames overlap. In this embodiment, an example is given where the areas in the waste pit 101, the first area, and the second area are identified by address.

[0087] Figure 5 shows an example of a flame detection result display screen 500. The flame detection result display screen 500 is displayed, for example, on a computer display, which is one of the output devices connected to the detection device 210.

[0088] In Figure 5, the flame detection result display screen 500 includes a first flame detection result display area 510, a second flame detection result display area 520, and a crane position display area 530.

[0089] The first flame detection result display area 510 and the second flame detection result display area 520 display information indicating the detection result of whether or not smoke and flames are generated.

[0090] Figure 5 illustrates a case where the first detection unit 212 detects that flames have occurred at addresses B5, B6, C5, and C6 based on the first trained model, and the confidence level of the inference by the first trained model is 0.70. Also in Figure 5, the first detection unit 212 detects that flames have occurred at addresses C5 and C6 based on the language model 221, and the confidence level of the inference by the language model 221 is 0.90. Also in Figure 5, the second detection unit 213 detects that flames have occurred at addresses C5 and D5 based on the infrared sensor 170. Also in Figure 5, the second detection unit 213 does not detect smoke or flames based on the ultraviolet sensor 160.

[0091] In the first flame detection result display area 510, information indicating the detection result of whether or not smoke and flames are generated is displayed using an image. Figure 5 illustrates a case where a simulated pit image 511, which simulates a plan view of the waste pit 110, is displayed in the first flame detection result display area 510. The simulated pit image 511 displays text information indicating each address (A1 to D9) assigned to the waste pit 110, as well as the boundary lines of each address. The simulated pit image 511 also displays flame detection areas 512a to 512c, overlap detection area 513, and crane position information 514a to 514b.

[0092] Flame detection areas 512a to 512c indicate the addresses where smoke or flames were detected. Figure 5 illustrates a case where flame detection areas 512a to 512c are displayed by surrounding the address where smoke or flames have been detected with a line of a different type than the boundary line. In this embodiment, the case is illustrated in which the address detected by which detection module can be identified by displaying flame detection areas 512a to 512c with different line types for each type of detection module. The detection module includes a trained model and an invisible light sensor. More specifically, in this embodiment, the detection module includes a first trained model, a language model 221, an ultraviolet sensor 160, and an infrared sensor 170. In Figure 5, flame detection areas 512a to 512b are an example of first area information, and flame detection area 512c is an example of second area information.

[0093] However, this is not necessarily required if the first area where the first detection unit 212 detected smoke / flames and the second area where the second detection unit 213 detected smoke / flames can be compared. For example, addresses detected by the ultraviolet sensor 160 and the infrared sensor 170 may be displayed with the same line type. Similarly, addresses detected by the language model 221 and the first trained model may be displayed with the same line type. In addition to or instead of using different line types for the lines displayed next to the boundary lines of addresses where smoke / flames are detected, the display form of the character information indicating the address (for example, at least one of color, thickness, and size) may be different.

[0094] The duplicate detection area 513 is information indicating the addresses that overlap (duplicate addresses) between the addresses detected by the first detection unit 212 and the addresses detected by the second detection unit 213.

[0095] Figure 5 illustrates a case in which, among the addresses detected by the first detection unit 212 and the addresses detected by the second detection unit 213, overlapping addresses are displayed in a way that makes them distinguishable from other addresses. This is achieved by displaying a pattern in addition to the character information indicating the address only in the area of ​​the addresses that overlap, while not displaying a pattern in the area of ​​the other addresses (the background of the character information indicating the address is white). The method for displaying these addresses in a distinguishable way is not limited to this method. For example, instead of or in addition to this, the display form of the character information indicating the address (for example, at least one of color, thickness, and size) may be varied.

[0096] Furthermore, the crane position information 514a and 514b indicate the current positions of cranes 120a and 120b (buckets 124a and 124b), respectively. Figure 5 illustrates the case where the crane position information 514a and 514b are displayed as white circles indicating the current positions of cranes 120a and 120b (buckets 124a and 124b). The current positions of cranes 120a and 120b (buckets 124a and 124b) are measured, for example, by using pulse generators installed on the traveling devices 121a to 121b and the traversing devices 122a to 122b, or GNSS (Global Navigation Satellite System) receivers installed on the traversing devices 122a to 122b. The measurement of the current positions of cranes 120a and 120b is achieved by known methods, so a detailed explanation is omitted here.

[0097] In the second flame detection result display area 520, information indicating the detection results of whether or not smoke and flames are generated is displayed in a table format. Figure 5 illustrates an example where the second flame detection result display area 520 displays a table having columns for detection module, detection result (confidence level), and detection area.

[0098] Figure 5 illustrates a case where the name of the detection module displayed in the detection module column is enclosed in a line type indicating the flame detection areas 512a to 512c. This makes it easier to understand which detection module the flame detection areas 512a to 512c represent. The information indicating which detection module the line type indicating the flame detection areas 512a to 512c represents may be displayed, for example, next to the first flame detection result display area 510. In Figure 5, among the information shown in the detection area column, the address information displayed in the row for the first trained model and language model, which is enclosed in the same line type as the flame detection areas 512a and 512b, is an example of first area information, and the address information displayed in the row for the infrared sensor, which is enclosed in the same line type as the flame detection area 512c, is an example of second area information.

[0099] Furthermore, Figure 5 illustrates an example of how, among the addresses displayed in the detection area column, the addresses detected by the first detection unit 212 and the addresses detected by the second detection unit 213 that overlap (C5) are enclosed in a rectangular frame, thereby making the overlapping addresses detected by the first detection unit 212 and the second detection unit 213 distinguishable from other addresses. The method for distinguishing these addresses is not limited to this method. For example, instead of or in addition to this, the display form of the character information indicating the address (for example, at least one of color, boldness, and size) may be varied.

[0100] In the crane position display area 530, information indicating the current positions of cranes 120a and 120b (buckets 124a and 124b) is displayed in a table format. Figure 5 illustrates a case where a table with columns for crane No., area, and location is displayed in the crane position display area 530. The crane No. column displays the identification information for cranes 120a and 120b. The area column displays the address of the current location of cranes 120a and 120b (buckets 124a and 124b). The location column displays the x and y coordinates of the current location of cranes 120a and 120b (buckets 124a and 124b). The x and y coordinates are coordinates where the origin is the lower left corner of the pit simulation image 511 displayed in the first flame detection result display area 510, with the positive x-axis pointing to the right of the screen and the positive y-axis pointing upwards. Figure 5 illustrates a case where the x and y coordinate values ​​"0, 5, ..., 20" and "0, 5, ..., 45" are displayed next to the pit simulation image 511.

[0101] <<<Second control process>>> Next, an example of the second control process will be described. The second control process, like the first control process, includes control processing related to notifying the results of smoke and flame detection in the waste pit 110. In the second control process, the detection results of the first detection unit 212 and the detection results of the second detection unit 213 are adaptively notified according to the conditions for notifying them.

[0102] When the mode information stored in the memory unit 215 indicates a first detection mode, the control unit 214 causes the first detection result information, indicating the detection result of the first detection unit 212, to be reported to the second notification unit when the first detection unit 212 detects the occurrence of smoke or flames. Furthermore, when the mode information stored in the memory unit 215 indicates a second detection mode, the control unit 214 causes the second detection result information, indicating the detection result of the second detection unit 213, to be reported to the second notification unit when the second detection unit 213 detects the occurrence of smoke or flames. Furthermore, when the mode information stored in the memory unit 215 indicates a third detection mode, the control unit 214 causes the second notification unit to report the third detection result information, indicating the detection results of the first detection unit 212 and the second detection unit 213, respectively, when the first detection unit 212 detects the occurrence of smoke or flames and the second detection unit 213 also detects the occurrence of smoke or flames. In the following explanation, when referring to the first detection mode, the second detection mode, and the third detection mode collectively, they will be referred to as "detection modes" as necessary.

[0103] The first detection mode, the second detection mode, and the third detection mode may be set, for example, by the operator of the detection device 210 (central control room). In this case, the operator of the detection device 210 may input detection mode information indicating one of the first detection mode, the second detection mode, and the third detection mode to the detection device 210 by operating an input device connected to the detection device 210. In this case, the control unit 214 may perform storage control to store the detection mode indicated in the detection mode information thus input in the storage unit 215 as detection mode information for the detection device 210.

[0104] Furthermore, the first detection mode, the second detection mode, and the third detection mode may be set, for example, by the operator of the external control device 230 (a facility that remotely monitors the operation of a waste treatment facility, etc.). In this case, when the communication unit 216 receives detection mode information transmitted from the external control device 230, the control unit 214 may store the detection mode indicated in the detection mode information in the storage unit 215 as detection mode information for the external control device 230.

[0105] Furthermore, the first detection mode, the second detection mode, and the third detection mode may be set, for example, by the operator of the processing unit 240 (crane control room). In this case, when the control unit 214 receives detection mode information transmitted from the processing unit 240 by the communication unit 216, it may store the detection mode indicated in the detection mode information in the storage unit 215 as detection mode information for the processing unit 240.

[0106] Furthermore, the first detection mode, the second detection mode, and the third detection mode may be set in common in the detection device 210 (central control room), the processing device 240 (crane operation room), and the external control device 230 (facility for remotely monitoring the operation of the waste treatment facility, etc.).

[0107] The second notification unit includes, for example, at least one of a display unit (e.g., a computer display), a light-emitting unit (e.g., a lamp), and a sound-emitting unit (e.g., a buzzer). The second notification unit may be a notification unit that is communicatively connected to the detection device 210, a notification unit that is communicatively connected to the external control device 230, or a notification unit that is communicatively connected to the processing device 240. Furthermore, the second notification unit may be the same notification unit as the first notification unit, or a different notification unit from the first notification unit.

[0108] <<<Third Control Process>>> Next, an example of the third control process will be described. The third control process includes the control process related to the detection instruction mentioned above in the section on the first detection unit 212 and the section on the sixth condition in the second detection unit 213.

[0109] When the first detection unit 212 or the second detection unit 213 detects the occurrence of smoke or flames based on either the learned model or the invisible light sensor, the control unit 214 causes at least one of the first detection unit 212 and the second detection unit 213 to perform smoke or flame detection processing based on at least one of the remaining sensors that have not detected the occurrence of smoke or flames. In this embodiment, an example is given where the learned model includes a language model 221 and a first learned model, and the invisible light sensor includes an ultraviolet sensor 160 and an infrared sensor 170. In this case, when the control unit 214 detects the occurrence of smoke or flames in the garbage pit 110 based on either the language model 221, the first learned model, the ultraviolet sensor 160, or the infrared sensor 170, it causes at least one of the remaining sensors to detect whether or not smoke or flames are occurring in the garbage pit 110. For example, if the first detection unit 212 detects smoke or flames based on the first trained model, the control unit 214 may issue a detection instruction to the first detection unit 212 to detect smoke or flames based on the language model 221. Alternatively, in addition to or instead of this detection instruction, the control unit 214 may issue a detection instruction to the second detection unit 213 to detect smoke or flames based on both or one of the ultraviolet sensor 160 and the infrared sensor 170. It may be predetermined which detection module should be used to detect smoke or flames when smoke or flames are detected based on which detection module.

[0110] Furthermore, the control unit 214 may cause the other detection unit 213 to perform smoke / flame detection processing when one of the first detection unit 212 or the second detection unit 213 detects the occurrence of smoke / flames. In this case, if the first detection unit 212 is the other detection unit, the control unit 214 may issue a detection instruction to the first detection unit 212 instructing it to detect smoke / flames based on all of the trained models (in this embodiment, the language model 221 and the first trained model), or it may issue a detection instruction to the first detection unit 212 instructing it to detect smoke / flames based on a part of the trained models (in this embodiment, the language model 221 or the first trained model). Similarly, if the second detection unit 213 is the other detection unit, a detection instruction may be given to the second detection unit 213 to instruct it to detect smoke and flames based on an invisible light sensor (in this embodiment, an ultraviolet sensor 160 and an infrared sensor 170), or a detection instruction may be given to the second detection unit 213 to instruct it to detect smoke and flames based on a part of the invisible light sensor (in this embodiment, the ultraviolet sensor 160 or the infrared sensor 170). The detection module on which smoke and flames are detected may be predetermined.

[0111] Furthermore, when the second detection unit 213 detects the occurrence of smoke or flames (when one of the aforementioned detection units is the second detection unit 213), the control unit 214 may use an enlarged version of the captured image data corresponding to the area in the garbage pit 110 where the second detection unit 213 detected the occurrence of smoke or flames as the estimated captured image data (estimated pit image data), and have the first detection unit 212 perform smoke or flame detection processing. In this way, the size of the estimated captured image data can be reduced. Therefore, it is possible to reduce the amount of processing required in the trained model and to narrow the estimation range in the trained model. In this case, the control unit 214 may use an enlarged version of the pit image taken at the closest shooting time to the shooting time of the ultraviolet image and / or infrared image, which includes the address corresponding to the pixel where the occurrence of smoke or flames was detected, as the estimated pit image in the second detection unit 213. The region to be clipped, which includes the address corresponding to the pixel where smoke or flame is detected, may be determined according to a predetermined logic. For example, the region to be clipped may be a rectangular region that includes the address corresponding to the pixel where smoke or flame is detected, and the addresses adjacent to that address.

[0112] <<<Fourth Control Process>>> Next, an example of the fourth control process will be described. The fourth control process includes a control process related to adjusting a predetermined threshold for the confidence level of the estimation results in the trained model. As described above in the section on the first detection unit 212, the first detection unit 212 may, for example, adopt the detection result output from the trained model as correct if the confidence level output from the trained model exceeds a predetermined threshold. The control unit 214 may lower the predetermined threshold if the first detection unit 212 does not detect smoke or flames, but the second detection unit 213 does. In this case, the control unit 214 changes the information of the predetermined threshold used by the first detection unit 212. The amount of change to the predetermined threshold may be predetermined. Alternatively, the amount of change to the predetermined threshold may be determined according to the number of times the predetermined threshold has been changed. For example, the absolute value of the amount of change to the predetermined threshold may be increased as the number of times the predetermined threshold has been changed increases.

[0113] <<<Fifth Control Process>>> Next, an example of the fifth control process will be explained. The fifth control process includes control processes related to water discharge. The control unit 214 controls the water discharge unit to discharge water onto the smoke or flame when both the first detection unit 212 and the second detection unit 213 detect smoke or flame. In this embodiment, an example is shown in which the control unit 214 controls the water discharge unit by outputting water discharge instruction information to the water discharge control device that controls the water discharge devices 150a to 150b, instructing it to discharge water. The water discharge instruction information may include information indicating an address as an example of the location where the smoke or flame is generated. In this case, the water discharge control device operates the water discharge device so that water is discharged from at least one of the water discharge devices 150a to 150b toward that address.

[0114] <<<6th Control Process>>> Next, an example of the sixth control process will be explained. The sixth control process, like the fifth control process, also includes control processes related to water discharge. Based on the first detection result information indicating the detection result of the first detection unit 212 and the second detection result information indicating the detection result of the second detection unit 213, the control unit 214 causes the determination unit to determine whether or not to control the water discharge unit to discharge water to the smoke and flames in the waste pit 110.

[0115] As described above in the section on the acquisition unit 211, in this embodiment, we illustrate a case in which a water discharge determination request prompt 410 is used as a prompt to cause a language model 221, which is an example of a determination unit, to determine whether or not to control the water discharge devices 150a to 150b, which are an example of a water discharge unit, to discharge water onto the smoke and flames based on the smoke and flame detection results in the garbage pit 110. In this case, the control unit 214 outputs to the input reception unit 217, for example, detection result information including first detection result information and second detection result information, and estimation image data (estimation pit image data) used by the first detection unit 212 to create the first detection result information. Based on these and the water discharge determination request prompt 410, the language model 221 determines whether or not to discharge water onto the garbage pit 110 and the address to which water should be discharged. The language model server 220 transmits water discharge determination result information indicating this determination result to the detection device 210. When the control unit 214 receives the water discharge determination result information from the input receiving unit 217, it generates and outputs water discharge instruction information based on the water discharge determination result information.

[0116] <<<Control Process 7>>> Next, an example of the seventh control process will be explained. The seventh control process, like the fifth control process, also includes control processes related to water discharge. For example, the control unit 214 instructs the communication unit 216 to transmit detection result information, including first detection result information indicating the detection result of the first detection unit 212 and second detection result information indicating the detection result of the second detection unit 213, to the external control device 230. After the first and second detection result information has been transmitted from the communication unit 216 to the external control device 230, when the communication unit 216 receives control information from the external control device 230 indicating whether or not to control the water discharge unit to discharge water onto the smoke and flames, (if the control information indicates that the water discharge unit should be controlled) it generates and outputs water discharge instruction information based on the control information. The control information may include information indicating an address as an example of information indicating the location to discharge water. Note that the information transmitted to the external control device 230 in the seventh control process is not limited to detection result information. For example, in addition to or instead of detection result information, estimation image data (estimation pit image data) indicating estimation images for which a judgment on the necessity of water discharge needs to be made may be transmitted to the external control device 230.

[0117] <<<Selection of Control Processes 5-7>>> The choice of which of the 5th to 7th control processes to implement as a control process related to water discharge may be determined, for example, depending on the water discharge mode. In this case, for example, when the mode information stored in the memory unit 215 indicates the first water discharge mode, the control unit 214 performs the fifth control process (discharging water at smoke and flames when both the first detection unit 212 and the second detection unit 213 detect smoke and flames).

[0118] Furthermore, for example, when the mode information stored in the memory unit 215 indicates the second water discharge mode, the control unit 214 performs a sixth control process (discharging water from the smoke and flames based on the determination result of the determination unit (language model server 220 in this embodiment)).

[0119] Furthermore, for example, when the mode information stored in the memory unit 215 indicates the third water discharge mode, the control unit 214 performs the seventh control process (discharging water at the smoke and flames based on the control information transmitted from the external control device 230).

[0120] The water discharge mode to be performed from among the first, second, and third water discharge modes may be set, for example, by the operator of the detection device 210 (central control room). Two or three of the first, second, and third water discharge modes may be set simultaneously. In this case, the operator of the detection device 210 may input water discharge mode information indicating one of the first, second, and third water discharge modes to the detection device 210 by operating an input device that is communicably connected to the detection device 210. Alternatively or in addition to this, the third water discharge mode may be set automatically during a predetermined period of time at night, for example, and the setting may be automatically canceled during other periods. Furthermore, when the communication unit 216 receives water discharge mode information indicating a water discharge mode specified in the external control device 230 or the processing device 240, the control unit 214 may set the water discharge mode indicated in the water discharge mode information. Furthermore, as mentioned above in the section on <garbage pit>, regardless of the settings of these water discharge modes, the water discharge devices 150a to 150b may be operated by workers.

[0121] <<<8th Control Process>>> Next, an example of the eighth control process will be described. The eighth control process includes control processing related to the transmission of smoke and flame detection results to the external control device 230.

[0122] If the first detection unit 212 or the second detection unit 213 detects smoke or flames after the water discharge unit has discharged water onto the smoke or flames, the control unit 214 instructs the communication unit 216 to transmit first detection result information, indicating the detection result of the first detection unit 212, and second detection result information, indicating the detection result of the second detection unit 213, to the external control device 230. As a result, the first and second detection result information are transmitted from the communication unit 216 to the external control device 230, and the external control device 230 can understand that smoke or flames have been detected despite the water discharge. In this case, for example, the external control device 230 may transmit control information to the detection device 210 indicating whether or not to control the water discharge unit to discharge water onto the smoke or flames. If the control information received by the communication unit 216 indicates that the water discharge unit should be controlled, the control unit 214 may generate and output water discharge instruction information based on that control information. Furthermore, the control unit 214 may instruct the communication unit 216 to transmit the estimation image data (estimation pit image data) used by the first detection unit 212 for creating the first detection result information to the external control device 230.

[0123] <<<9th Control Process>>> Next, an example of the ninth control process will be described. The ninth control process includes the control process related to the detection instruction mentioned above in the section on the <<<fourth condition>>> of the section on the <<first detection unit 212>>. For the sake of explanation, the ninth control process will be described here as a separate control process from the third control process, but the ninth control process may also be an example of one aspect of the third control process mentioned above.

[0124] When the communication unit 216 receives estimation instruction information from the external control device 230 instructing the language model 221 to estimate whether or not smoke and flames are present, the control unit 214 issues a detection instruction to the first detection unit 212 instructing it to detect smoke and flames based on the language model 221.

[0125] <<<10th Control Process>>> Next, an example of the tenth control process will be described. The tenth control process is a control process related to retraining the trained model. The control unit 214 performs a storage control process to store in the storage unit 215 the estimated captured image data (estimated pit image data) acquired by the acquisition unit 211, corresponding to the case where the first detection unit 212 does not detect the occurrence of smoke or flames, but the second detection unit 213 does detect the occurrence of smoke or flames. The estimated captured image data stored in the storage unit 215 in this way is, for example, the estimated captured image data (estimated pit image data) with the closest shooting time to the shooting time of the ultraviolet image and / or infrared image used when the second detection unit 213 detected the occurrence of smoke or flames. In this case, the first detection unit 212 does not detect smoke or flames based on the said estimated captured image data (estimated pit image data). The control unit 214 may use the estimated captured image data stored in the storage unit 215 in this way to retrain the first trained model used by the first detection unit 212. The control unit 214 may also use the estimated captured image data stored in the storage unit 215 in this way to fine-tune the language model 221.

[0126] <<Storage section 215>> The memory unit 215 stores various types of information necessary for the detection device 210 to perform processing. For example, the mode information (information indicating the 1st to 3rd detection modes) mentioned above in the section of <<<2nd control processing>>>, the mode information (information indicating the 1st to 3rd water discharge modes) mentioned above in the section of <<<5th to 7th control processing selection>>>, and the estimated captured image data (estimated pit image data) mentioned above in the section of <<<10th control processing>>> are stored in the storage unit 215.

[0127] <<Communications Department 216>> The communication unit 216 communicates with the detection device 210 and an external device. In this embodiment, an example is given in which the communication unit 216 communicates with a device including an external control device 230 and a processing device 240. For example, as described above in the sections on the first control process, the seventh control process, and the eighth control process, the communication unit 216 transmits detection result information (first detection result information, second detection result information) indicating the detection results of the first detection unit 212 and the second detection unit 213 to the external control device 230 and the processing device 240.

[0128] Furthermore, as mentioned above in the section on the second control process, the communication unit 216 receives detection mode information (information indicating at least one of the first to third detection modes) transmitted from the external control device 230 and the processing device 240.

[0129] Furthermore, as described above in sections such as <<<7th control process>>> and <<<8th control process>>>, the communication unit 216 receives control information transmitted from the external control device 230 (information indicating whether or not to control the water discharge unit to spray water on smoke and flames).

[0130] Furthermore, as mentioned above in the section on the <<<9th control process>>>, the communication unit 216 receives estimation instruction information (information instructing the language model 221 to estimate whether or not smoke and flames are present) transmitted from the external control device 230.

[0131] <<Input reception unit 217>> The input receiving unit 217 inputs and receives (sends and receives) various types of information to and from the language model server 220. The input receiving unit 217 may also input and receive information to and from the language model 221 via an API (Application Programming Interface) or the like.

[0132] For example, the input receiving unit 217 inputs the detection prompt 400 and the water discharge determination request prompt 410, as described in the section on the acquisition unit 211, to the language model server 220 (language model 221).

[0133] Furthermore, for example, the input receiving unit 217 inputs the estimation pit image data output from the first detection unit 212 to the language model server 220 (language model 221) as described above in the sections <<<First Condition>> to <<<Fourth Condition>>> of the <<First Detection Unit 212>> section. In this case, the input receiving unit 217 receives the detection result information output (transmitted) from the language model server 220 (language model 221) as detection result information based on the estimation pit image data. This detection result information indicates the detection result of smoke and flame generation by the language model 221. Based on this detection result information, first detection result information indicating the detection result of the first detection unit 212 is created.

[0134] Furthermore, for example, the input receiving unit 217 inputs the detection result information and estimated captured image data (estimated pit image data) output from the control unit 214 to the language model server 220 (language model 221) as described above in the section on the sixth control process of the control unit 214. This detection result information is used by the language model 221 to make a determination, including whether or not water discharge is necessary. In this case, the input receiving unit 217 receives the water discharge determination result information output (transmitted) from the language model server 220 (language model 221) as water discharge determination result information based on the detection result information and estimated pit image data.

[0135] For the sake of explanation, here we have illustrated the case where the communication unit 216 handles the exchange of information with the processing unit 240 and the external control unit 230, the input receiving unit 217 handles the exchange of information with the language model server 220, and the acquisition unit 211 handles the acquisition of other information. However, the functions of the acquisition unit 211, the communication unit 216, and the input receiving unit 217 may be implemented in a single module or in multiple modules.

[0136] <Detection Method> Next, an example of a detection method performed using the detection device 210 will be described with reference to the flowcharts in Figures 6-1 to 6-4. For the sake of simplicity, it will be assumed that the detection prompt 400 and the water discharge judgment request prompt 410 have already been input into the language model 221 before the processing of the flowcharts in Figures 6-1 to 6-4 begins. However, this is not necessarily required. For example, when inputting the estimation pit image data into the language model 221 in step S613, the identification prompt 400 may be input into the language model 221 each time. Also, for example, when requesting a water discharge judgment in step S646, the water discharge judgment request prompt 410 may be input into the language model 221 each time. Furthermore, it will be assumed that the detection device 210 is set to be readable for the information that needs to be pre-configured in order to execute the processing of the flowcharts in Figures 6-1 to 6-4 (for example, detection mode, water discharge mode, and various predetermined thresholds).

[0137] In step S601, the acquisition unit 211 acquires pit-in image data captured by the imaging devices 140a to 140b. For simplicity, this example illustrates the case where pit-in image data representing one pit-in image (or a frame image if it is a moving image) is acquired in one step S601. The pit-in image data acquired in step S601 is, for example, data representing the most recent image captured by the imaging devices 140a to 140b. If multiple pit-in image data representing multiple pit-in images are acquired in one step S601, for example, the processing described in the following steps is performed in parallel for each of the multiple pit-in image data in the subsequent steps.

[0138] Next, in step S602, the acquisition unit 211 acquires ultraviolet sensor data and infrared sensor data. For the sake of simplicity, here we will illustrate a case where, in one step S602, ultraviolet image data representing one ultraviolet image and infrared image data representing one infrared image are acquired. Note that the ultraviolet image data and infrared image data acquired in step S602 are, for example, the most recent image data.

[0139] Next, in step S603, the acquisition unit 211 determines whether or not it is the timing for smoke / flame detection by the second detection unit 213. The timing for smoke / flame detection by the second detection unit 213 occurs, for example, at a preset period. Step S603 corresponds to the process of determining whether or not the seventh condition described above is met. If, as a result of this determination, it is not the timing for smoke / flame detection by the second detection unit 213 (NO in step S603), the process of step S609, described later, is performed. On the other hand, if it is the timing for smoke / flame detection by the second detection unit 213 (YES in step S603), the process of step S604 is performed.

[0140] In step S604, the second detection unit 213 detects the presence or absence of smoke and flames in the waste pit 110 based on the ultraviolet sensor data acquired in step S602. The second detection unit 213 also detects the presence or absence of smoke and flames in the waste pit 110 based on the infrared sensor data acquired in step S602. In this embodiment, if the second detection unit 213 detects the presence or absence of smoke and flames, it detects the address as an example of the location where the smoke and flames were detected. In this case, the second detection unit 213 uses information indicating the presence or absence of smoke and flames and information indicating the address where the smoke and flames were detected as the second detection result information.

[0141] Next, in step S605, the control unit 214 determines whether the second detection unit 213 detected smoke or flames in step S604. If the second detection unit 213 detects smoke or flames (i.e., YES in step S605), the process in step S606 is performed. This step S605 corresponds to one of the processes for determining whether the first condition described above is met.

[0142] In step S606, the control unit 214 creates estimated pit image data by enlarging the image data from the pit image data acquired in step S602 that corresponds to the area where the second detection unit 213 detected the generation of smoke and flames. This step S606 is the process corresponding to the third control process described above. Then, the process of step S608, which will be described later, is performed.

[0143] On the other hand, if the second detection unit 213 does not detect smoke or flames in step S605 (the result is NO in step S605), the process in step S607 is performed. In step S607, the control unit 214 uses the pit image data acquired in step S602 as the estimated pit image data. Then, the process in step S608 is performed.

[0144] In step S608, the control unit 214 issues a detection instruction to the first detection unit 212 to detect smoke and flames based on the first trained model. The first detection unit 212 detects the presence or absence of smoke and flames in the waste pit 110 based on the estimation pit image data obtained in step S606 or 607 and the first trained model, and creates first detection result information indicating the detection result. The first detection result information includes, for example, information indicating the presence or absence of smoke and flames, information indicating the presence or absence of smoke and flames at each address, information indicating the reliability of the detection result for the presence or absence of smoke and flames at each address, information indicating whether smoke or flames were detected at each address, and information indicating the type of waste at each address. In this embodiment, an example is given in which the first detection unit 212 adopts the detection result output from the first trained model when the reliability output from the first trained model exceeds a predetermined threshold. Step S608 is a process corresponding to the third control process described above. Once the process in step S608 is completed, the process in step S611, shown in Figure 6-2 below, is performed.

[0145] As mentioned above, in step S603, if it is not the timing for smoke / flame detection by the second detection unit 213 (if the result is NO in step S601), the process in step S609 is performed. In step S609, the acquisition unit 211 determines whether or not it is the timing for smoke / flame detection by the first detection unit 212. The timing for smoke / flame detection by the first detection unit 212 occurs, for example, at a preset period. Step S609 corresponds to the process of determining whether or not the fifth condition mentioned above is met. As a result of this determination, if it is the timing for smoke / flame detection by the first detection unit 212 (if the result is YES in step S609), the process in step S607 is performed. On the other hand, if it is not the timing for smoke / flame detection by the first detection unit 212 (if the result is NO in step S609), the process in step S601 is performed.

[0146] As mentioned above, once the processing in step S608 is completed, the processing in step S611 shown in Figure 6-2 is performed. In step S611, the control unit 214 determines whether or not smoke or flames were detected in step S604 or step S608. Step S611 corresponds to one of the processes that determines whether or not the first condition described above is met. If, as a result of this determination, smoke or flames are not detected (NO in step S611), the processing in step S614, which will be described later, is performed. On the other hand, if smoke or flames are detected (YES in step S611), the processing in step S612 is performed.

[0147] In step S612, the control unit 214 determines whether the confidence level calculated in step S608 when detecting smoke and flames is within a predetermined range (greater than or equal to a predetermined value and less than or equal to a predetermined value). If the confidence level calculated in step S608 is not within the predetermined range (NO in step S612), the process in step S615, described later, is performed. On the other hand, if the confidence level calculated in step S608 is within the predetermined range (YES in step S612), the process in step S613 is performed. Step S612 corresponds to the process of determining whether the second condition described above is met.

[0148] In step S613, the control unit 214 issues a detection instruction to the first detection unit 212 to detect smoke and flames based on the language model 221. At this time, the first detection unit 212 may add information indicating the smoke and flame detection result based on the first trained model and / or invisible light sensor (for example, the location where smoke and flames were detected (in this embodiment, the address)) to the estimation pit image data. The first detection unit 212 outputs the estimation pit image data (and additional information) obtained in step S606 or 607 to the input receiving unit 217. The input receiving unit 217 inputs the estimation pit image data to the language model server 220 (language model 221). The first detection unit 212 then obtains detection result information from the language model server 220 (language model 221) via the input receiving unit 217. The detection result information from the language model server 220 (language model 221) includes, for example, information indicating whether or not smoke and flames are present, information indicating whether or not smoke and flames are present at each address, information indicating the reliability of the detection result for the presence or absence of smoke and flames at each address, information indicating whether or not smoke and flames were detected at each address, and information indicating the type of waste at each address. In this embodiment, an example is given in which the first detection unit 212 adopts the detection result output from the language model 221 when the reliability included in the detection result information output from the language model 221 exceeds a predetermined threshold.

[0149] The first detection unit 212 creates first detection result information indicating the detection result of smoke and flame generation by the language model 221 based on the detection result information from the language model server 220 (language model 221). The first detection result information includes, for example, information indicating whether or not smoke and flame generation has occurred, information indicating whether or not smoke and flame generation has occurred at each address, information indicating the reliability of the detection result of whether or not smoke and flame generation has occurred at each address, and information indicating whether or not smoke and flame were detected at each address. Then, the processing of step S617, which will be described later, is performed. This step S613 is the processing corresponding to the third control processing described above.

[0150] As mentioned above, if no smoke or flame is detected in step S611 (the result is NO in step S611), the process in step S614 is performed. In step S614, the control unit 214 determines whether the confidence level calculated in step S608 when smoke or flame was detected is within a predetermined range (greater than or equal to a predetermined value and less than or equal to a predetermined value). If the confidence level calculated in step S608 is within the predetermined range (the result is YES in step S614), the process in step S613 described above is performed. On the other hand, if the confidence level calculated in step S608 is not within the predetermined range (the result is NO in step S614), the process in step S615 is performed. Step S614 corresponds to the process of determining whether the second condition described above is met.

[0151] As mentioned above, if the result in step S612 or S614 is NO, the process in step S615 is performed. In step S615, the control unit 214 determines whether the communication unit 216 has received estimation instruction information from the external control device 230 instructing the language model 221 to estimate whether or not smoke and flames are present. If the estimation instruction information has been received from the external control device 230 (if the result in step S615 is YES), the process in step S613 described above is performed. Thus, step S613, which is performed when the result in step S615 is YES, corresponds to the ninth control process described above. On the other hand, if the estimation instruction information has not been received from the external control device 230 (if the result in step S615 is NO), the process in step S616 is performed.

[0152] In step S616, the control unit 214 determines whether water discharge instruction information has been output (i.e., water has been discharged in response to smoke and flames in the waste pit 110) and whether the occurrence of smoke and flames has been detected by at least one of the first detection unit 212 and the second detection unit 213 (i.e., whether at least one of the preceding steps S605 and S611 of step S616 was determined to be YES). The control unit 214 determines, for example, whether water discharge instruction information has been output at a time predetermined before the timing of executing step S614. This predetermined time may be set in advance based on, for example, the time expected to be required for the smoke and flames to disappear due to water discharge. Step S616 corresponds to the process of determining whether the third condition described above is met.

[0153] If, as a result of this determination, water discharge instruction information has been output and smoke or flames are detected in at least one of the first detection unit 212 and the second detection unit 213 (if YES in step S616), the process in step S613 described above is performed. On the other hand, if water discharge instruction information has not been output, or if smoke or flames are not detected in at least one of the first detection unit 212 and the second detection unit 213 (if NO in step S616), the process in step S617 is performed.

[0154] Steps S612, S614, S615, and S616 correspond to the process of determining whether the first condition described above is met. For the sake of simplicity, the flowcharts in Figures 6-1 to 6-4 illustrate a case where the process of determining whether the sixth condition is met is not performed. However, the process described in step S604 may be performed based on the detection instruction from the control unit 214.

[0155] As mentioned above, if the processing in step S613 is completed, or if NO is determined in step S616, the processing in step S617 is performed. In step S617, the control unit 214 determines whether or not it is necessary to change a predetermined threshold for the reliability of the smoke / flame detection result in the trained model. Here, we illustrate the case where it is determined that it is necessary to change the predetermined threshold for the reliability of the smoke / flame detection result in the trained model when YES is determined in step S605 and NO is determined in step S611. Also, here we illustrate the case where it is determined that it is necessary to change the predetermined threshold for the reliability of the smoke / flame detection result in the trained model when YES is determined in step S605 and the occurrence of smoke / flame is not detected in step S613.

[0156] If, as a result of this determination, it is determined that there is no need to change the predetermined threshold for the confidence level of the smoke / flame detection results in the trained model (the answer is NO in step S617), then the processes in steps S618 and S619 are omitted, and the process in step S621, shown in Figure 6-3 later, is performed. On the other hand, if it is necessary to change the predetermined threshold for the confidence level of the smoke / flame detection results in the trained model (the answer is YES in step S617), then the process in step S618 is performed.

[0157] In step S618, the control unit 214 lowers a predetermined threshold for the reliability of the smoke / flame detection result in the trained model. If it is determined to be YES in step S605 and NO in step S611, the control unit 214 lowers a predetermined threshold for the reliability of the smoke / flame detection result in the first trained model. Also, if it is determined to be YES in step S605 and no smoke or flame is detected in step S613, the control unit 214 lowers a predetermined threshold for the reliability of the smoke / flame detection result in the language model 221. This step S618 corresponds to the fourth control process described above. Then, the process in step S619 is performed.

[0158] In step S619, the control unit 214 stores the pit image data acquired in step S602 in the storage unit 215. Step S619 corresponds to the 10th control process (storage control process) described above. When the process in step S619 is completed, the process in step S621 shown in Figure 6-3 is performed.

[0159] In step S621, the control unit 214 determines whether or not water discharge instruction information has been output (i.e., water has been discharged against the smoke and flames in the waste pit 110). The control unit 214 determines whether or not water discharge instruction information has been output at a predetermined time before the timing of executing step S621. The predetermined time may be set in advance based on, for example, the time expected to be required for the smoke and flames to disappear by water discharge. If, as a result of this determination, water discharge instruction information has not been output (the result is NO in step S621), the processing in steps S622 to S623 is omitted, and the processing in step S624, which will be described later, is performed.

[0160] On the other hand, if the water discharge instruction information has already been output (if the answer is YES in step S621), the process in step S622 is performed. In step S622, the control unit 214 determines whether at least one of the first detection unit 212 and the second detection unit 213 has detected the generation of smoke or flames. The first detection unit 212 detects the generation of smoke or flames if at least one of the following is determined: in step S611, YES is determined; or in step S613, the generation of smoke or flames is detected. The second detection unit 213 detects the generation of smoke or flames if YES is determined in step S611.

[0161] If, as a result of this determination, neither the first detection unit 212 nor the second detection unit 213 detects the generation of smoke or flames (the result is NO in step S622), the process in step S623 is omitted, and the process in step S624, described later, is performed. On the other hand, if at least one of the first detection unit 212 or the second detection unit 213 detects the generation of smoke or flames (the result is YES in step S622), the process in step S623 is performed.

[0162] In step S623, the control unit 214 instructs the communication unit 216 to transmit first detection result information, which indicates the detection result of the first detection unit 212, and second detection result information, which indicates the detection result of the second detection unit 213, to the external control device 230. The communication unit 216 then transmits the first detection result information and the second detection result information to the external control device 230. Step S623 corresponds to the eighth control process described above. After the processing in step S623 is completed, the processing in step S624 is performed.

[0163] In step S624, the control unit 214 selects one of the unselected devices from the detection device 210 (central control room), the processing device 240 (crane operation room), and the external control device 230. For example, the control unit 214 may select the devices one by one in the order of detection device 210, processing device 240, and external control device 230.

[0164] Next, in step S625, the control unit 214 reads the detection mode stored in the memory unit 215 that is set for the device selected in step S624. For simplicity of explanation, this example illustrates the case where a detection mode is set for each of the detection device 210, the processing unit 240, and the external control device 230. However, for example, there may be a device among the detection device 210, the processing unit 240, and the external control device 230 for which no detection mode is set. In this case, the device among the detection device 210, the processing unit 240, and the external control device 230 for which no detection mode is set may be selected in step S624. In this case, for example, in step S625, the control unit 214 cannot read the detection mode set for that device. In this case, the control unit 214 may perform the process of step S624 again without performing the process of step S626 (i.e., it may select a different device in step S624).

[0165] Next, in step S626, the control unit 214 determines the detection mode read in step S625. As described above in the section on the <<<second control processing>>> of the <<control unit 214>>, in this embodiment, the case where the detection mode is one of the first to third detection modes is given as an example. If the result of the determination in step S626 is that the detection mode is the first detection mode, the processing in step S627 is performed. The first detection mode is a mode in which the first detection unit 212 notifies first detection result information indicating the detection result of the first detection unit 212 when it detects the generation of smoke or flames. In the first detection mode, second detection result information indicating the detection result of the second detection unit 213 is not notified. The processing when the detection mode is the second detection mode or the third detection mode (steps S630 to S631, steps S632 to S633) will be described later.

[0166] In step S627, the control unit 214 determines whether the first detection unit 212 has detected the generation of smoke or flames. If the first detection unit 212 has not detected the generation of smoke or flames (the result in NO in step S627), the process in step S627 is omitted and the process in step S629, which will be described later, is performed. On the other hand, if the first detection unit 212 has detected the generation of smoke or flames (the result in YES in step S627), the process in step S628 is performed.

[0167] In step S628, the control unit 214 causes the notification unit connected to the device selected in step S624 to notify the notification unit of the first detection result information, which indicates the detection result of the first detection unit 212. For example, the information obtained by removing the information based on the detection result of the second detection unit 213 from the flame detection result display screen 500 illustrated in Figure 5 is displayed on a computer display that is communicably connected to the device selected in step S624. In step S628, the flame detection area 512c based on the infrared sensor 170 is not displayed in the first flame detection result display area 510. Also, the overlap detection area 513 is not displayed because it shows the addresses that overlap (duplicate addresses) between the addresses detected by the first detection unit 212 and the addresses detected by the second detection unit 213. In addition, the rows for the infrared sensor and ultraviolet sensor are not displayed in the second flame detection result display area 520.

[0168] Next, in step S629, the control unit 214 determines whether all of the detection device 210, processing device 240, and external control device 230 were selected in step S624. If all of the detection device 210, processing device 240, and external control device 230 are selected (YES in step S629), the process in step S641 shown in Figure 6-4, which will be described later, is performed. On the other hand, if all of the detection device 210, processing device 240, and external control device 230 are not selected (NO in step S629), the process in step S624 described above is performed again. Then, in step S624, the control unit 214 selects one of the unselected devices from the detection device 210, processing device 240, and external control device 230, and performs the process from step S625 onwards for the selected device.

[0169] If the result of the determination in step S626 is that the detection mode is the second detection mode, the process in step S630 is performed. The second detection mode is a mode in which, when the second detection unit 213 detects the generation of smoke or flames, second detection result information indicating the detection result of the second detection unit 213 is not notified. In the second detection mode, first detection result information indicating the detection result of the first detection unit 212 is not notified.

[0170] In step S630, the control unit 214 determines whether the second detection unit 213 has detected the generation of smoke or flames. If the second detection unit 213 has not detected the generation of smoke or flames (the result in step S630 is NO), the process in step S631 is omitted and the process in step S629 described above is performed. On the other hand, if the second detection unit 213 has detected the generation of smoke or flames (the result in step S630 is YES), the process in step S631 is performed.

[0171] In step S631, the control unit 214 causes the notification unit, which is communicably connected to the device selected in step S624, to notify the notification unit of the second detection result information, which indicates the detection result of the second detection unit 213. For example, in the flame detection result display screen 500 illustrated in Figure 5, the information excluding the information based on the detection result of the first detection unit 212 is displayed on the computer display connected to the device selected in step S624. In step S631, the flame detection area 512a based on the first trained model and the flame detection area 512b based on the language model 221 are not displayed in the first flame detection result display area 510. The duplicate detection area 513 is also not displayed. In addition, the rows for the first trained model and the language model are not displayed in the second flame detection result display area 520. When the processing of step S631 is completed, the processing of step S641, shown in Figure 6-4, which will be described later, is performed.

[0172] If the result of the determination in step S626 above indicates that the detection mode is the third detection mode, the process in step S632 is performed. The third detection mode is a mode in which, when the first detection unit 212 detects the generation of smoke or flames and the second detection unit 213 also detects the generation of smoke or flames, third detection result information indicating the detection results of the first detection unit 212 and the second detection unit 213 is notified. In the third detection mode, both the first detection result information indicating the detection result of the first detection unit 212 and the second detection result information indicating the detection result of the second detection unit 213 (i.e., the third detection result information mentioned above) are notified.

[0173] In step S632, the control unit 214 determines whether the first detection unit 212 and the second detection unit 213 (both) have detected the generation of smoke or flames. If, as a result of this determination, the first detection unit 212 and the second detection unit 213 have not detected the generation of smoke or flames (the result in NO in step S632), the process in step S633 is omitted and the process in step S629 described above is performed. On the other hand, if the first detection unit 212 and the second detection unit 213 have detected the generation of smoke or flames (the result in YES in step S632), the process in step S633 is performed.

[0174] In step S633, the control unit 214 causes the notification unit, which is communicatively connected to the device selected in step S624, to notify the notification unit of the third detection result information, which indicates the detection results of the first detection unit 212 and the second detection unit 213, respectively. For example, as shown in the flame detection result display screen 500 illustrated in Figure 5, information indicating the detection result of the first detection unit 212 and information indicating the detection result of the second detection unit 213 are displayed on a computer display which is communicatively connected to the device selected in step S624. In step S633, the overlap detection area 513 may also be displayed. When the processing in step S633 is completed, the processing in step S629 described above is performed. Steps S624 to S629 are processes corresponding to the first control process and the second control process described above.

[0175] As mentioned above, if it is determined in step S629 that all of the detection device 210, processing device 240, and external control device 230 have been selected (if the result in step S629 is YES), then the process in step S641 shown in Figure 6-4 is performed. In step S641, the control unit 214 reads the water discharge mode stored in the storage unit 215.

[0176] Next, in step S642, the control unit 214 determines whether the water discharge mode read in step S641 is the first water discharge mode. As described above in the section on <<<Selection of the 5th to 7th control processes>>> in the section on <<Control Unit 214>>, in this embodiment, the case where the water discharge mode is the 1st to 3rd water discharge modes is illustrated. In the first water discharge mode, the 5th control process (discharging water at smoke / flames when both the first detection unit 212 and the second detection unit 213 detect smoke / flames) is performed. If the determination in step S642 is that the water discharge mode is not the first water discharge mode (if NO is obtained in step S642), the processes in steps S643 to S644 are omitted and the process in step S645, which will be described later, is performed. On the other hand, if the water discharge mode is the first water discharge mode (if YES is obtained in step S642), the process in step S643 is performed.

[0177] In step S643, the control unit 214 determines whether the first detection unit 212 and the second detection unit 213 (both) have detected the generation of smoke or flames. If, as a result of this determination, the first detection unit 212 and the second detection unit 213 have not detected the generation of smoke or flames (the result is NO in step S643), the processing in step S644 is omitted and the processing in step S645, which will be described later, is performed. On the other hand, if the first detection unit 212 and the second detection unit 213 have detected the generation of smoke or flames (the result is YES in step S643), the processing in step S644 is performed.

[0178] In step S644, the control unit 214 outputs water discharge instruction information to the water discharge control device that controls the water discharge devices 150a to 150b, instructing it to discharge water. As described above in the sections on the first detection unit 212 and the second detection unit 213, in this embodiment, an example of a smoke / flame generation location is given where an address is detected. In this case, it is preferable for the control unit 214 to include information indicating the address where the smoke / flame generation was detected in the water discharge instruction information. By doing so, the water discharge control device can select a water discharge device from among the water discharge devices 150a to 150b that is close to the address included in the water discharge instruction information, and operate the water discharge device so that water is discharged from the selected water discharge device toward that address. This reduces the range and amount of water discharged into the garbage pit 110. Steps S642 to S644 described above correspond to the fifth control process.

[0179] If the processing in step S644 is completed, or if the result in steps S642 and S643 is determined to be NO, the processing in step S645 is performed. In step S645, the control unit 214 determines whether the water discharge mode read in step S641 is the second water discharge mode. In the second water discharge mode, the sixth control process (discharging water onto the smoke and flames based on the determination result of the determination unit (language model server 220 in this embodiment)) is performed. If the result of the determination in step S645 is that the water discharge mode is not the second water discharge mode (if the result in step S645 is NO), the processing in steps S646 to S648 is omitted and the processing in step S649, which will be described later, is performed. On the other hand, if the water discharge mode is the second water discharge mode (if the result in step S645 is YES), the processing in step S646 is performed.

[0180] In step S646, the control unit 214 requests the language model 221 to make a water discharge decision by outputting detection result information, which includes first detection result information indicating the detection result of the first detection unit 212 and second detection result information indicating the detection result of the second detection unit 213, and estimation pit image data used by the first detection unit 212 to create the first detection result information, to the input reception unit 217. Based on these and the water discharge decision request prompt 410, the language model 221 decides whether or not to discharge water to the garbage pit 110 and the address to which water should be discharged. The language model server 220 transmits water discharge decision result information indicating this decision result to the detection device 210. The control unit 214 acquires the water discharge decision result information via the input reception unit 217.

[0181] Next, in step S647, the control unit 214 determines whether or not it is necessary to discharge water to the waste pit 110 based on the water discharge determination result information obtained in step S646. If the result of this determination is that water discharge is not necessary (NO in step S647), the process in step S648 is omitted and the process in step S649, which will be described later, is performed. On the other hand, if water discharge is necessary (YES in step S647), the process in step S648 is performed.

[0182] In step S648, the control unit 214 outputs water discharge instruction information to the water discharge control device that controls the water discharge devices 150a to 150b, based on the water discharge determination result information acquired in step S646. In this case as well, as described above in the explanation of the process in step S644, it is preferable that the control unit 214 include information indicating the address where the occurrence of smoke and flames was detected in the water discharge instruction information. Steps S645 to S648 described above correspond to the sixth control process.

[0183] If the processing in step S648 is completed, or if the result in steps S645 and S647 is NO, the processing in step S649 is performed. In step S649, the control unit 214 determines whether the water discharge mode read in step S641 is the third water discharge mode. In the third water discharge mode, the seventh control process (discharging water at the smoke and flames based on the control information transmitted from the external control device 230) is performed. If the result of the determination in step S649 is that the water discharge mode is not the third water discharge mode (if the result in step S649 is NO), the processing in step S653, which will be described later, is performed. On the other hand, if the water discharge mode is the third water discharge mode (if the result in step S649 is YES), the processing in step S650 is performed.

[0184] In step S650, the control unit 214 instructs the communication unit 216 to transmit detection result information, including the first detection result information and the second detection result information, to the external control device 230. The communication unit 216 then transmits the detection result information to the external control device 230.

[0185] Next, in step S651, the control unit 214 determines whether the communication unit 216 has received control information from the external control device 230 indicating whether or not to control the water discharge unit to discharge water onto the smoke and flames. This determination may be repeated, for example, until a predetermined time has elapsed since the instruction to transmit the detection result information in step S650. This predetermined time may be set in advance, for example, based on the time required for the external control device 230 to determine whether or not water discharge is necessary. If, as a result of this determination, control information has not been received (NO in step S651), the process in step S653, which will be described later, is performed. On the other hand, if control information has been received (YES in step S651), the process in step S652 is performed.

[0186] In step S652, the control unit 214 outputs water discharge instruction information to the water discharge control device that controls the water discharge devices 150a to 150b, based on the control information acquired in step S651. For example, if the control information includes information indicating an address as an example of information indicating the location to discharge water, it is preferable for the control unit 214 to include the information indicating the address in the water discharge instruction information. Steps S649 to S652 described above correspond to the seventh control process. Once the process in step S652 is completed, the process shown in the flowchart in Figure 6-4 is finished.

[0187] As described above, if the determination in step S649 indicates that the water discharge mode is not the third water discharge mode (i.e., NO in step S649), the process in step S653 is performed. In step S653, the control unit 214 determines whether or not a manual water discharge instruction has been given. For example, the control unit 214 determines whether or not the worker in charge of water discharge has performed a predetermined operation to operate the water discharge devices 150a to 150b. This predetermined operation may be performed on an input device connected to the detection device 210, or on an operating means such as an operation button separate from the detection device 210. In the latter case, for example, the control unit 214 may determine that a manual water discharge instruction has been given when the detection device 210 (communication unit 216) receives information indicating that a predetermined operation has been performed on the operating means. Note that the determination in step S653 may be performed by a device other than the detection device 210.

[0188] If, as a result of this determination, no manual water discharge instruction has been given (the result is NO in step S653), the process according to the flowchart in Figure 6-4 is terminated. On the other hand, if a manual water discharge instruction has been given, the process in step S654 is performed. In step S654, the control unit 214 outputs water discharge instruction information to the water discharge control device that controls the water discharge devices 150a to 150b, based on the operator's operation of the operating device that operates the water discharge devices 150a to 150b. The water discharge in step S654 corresponds to water discharge performed manually by the operator. Note that the process in step S654 may be performed by a device other than the detection device 210.

[0189] <Summary> As described above, in this embodiment, the detection device 210 detects the occurrence of smoke and flames in the waste pit 110 based on estimation image data showing an image of the waste pit 110 where waste is stored, and a trained model, and also detects the occurrence of smoke and flames in the waste pit 110 using an invisible light sensor. Therefore, for example, the advantages of using a trained model and the advantages of using an invisible light sensor can be enjoyed. The invisible light sensor includes, for example, at least one of an ultraviolet sensor 160 and an infrared sensor 170. Some invisible light sensors, such as the ultraviolet sensor 160 and the infrared sensor 170, have high reliability in their detection results. Also, some trained models can output a variety of detection results. Furthermore, when using an invisible light sensor, training is not required. Also, trained models have a short detection time. For the reasons exemplified above, the accuracy of smoke and flame detection in the waste pit 110 can be improved.

[0190] Furthermore, in this embodiment, when at least one of the first detection unit 212 and the second detection unit 213 detects the occurrence of smoke or flames, the detection device 210 causes the first notification unit to notify the detection result information indicating the respective detection results of the first detection unit 212 and the second detection unit 213. Therefore, information indicating the detection result of the occurrence (presence or absence) of smoke or flames when using both the learned model and the invisible light sensor can be provided to, for example, the central control room, the crane operation room, and at least one operator of the external control device.

[0191] Furthermore, in this embodiment, the detection device 210 displays on a display unit in a comparable manner first area information, which indicates a first area in the waste pit 101 where the first detection unit 212 detected smoke and flames, and second area information, which indicates a second area in the waste pit 101 where the second detection unit 213 detected smoke and flames. For example, a flame detection result display screen 500, which displays a first flame detection result display area 510 that overlays flame detection areas 512a to 512c on a simulated pit image 511, and a second flame detection result display area 520 that displays the areas where smoke and flames were detected in a table format, is displayed on a display unit installed in, for example, the central control room, the crane operation room, and the external control device. Therefore, at least one operator in the central control room, the crane operation room, and the external control device can compare the area where the first detection unit 212 detected smoke and flames with the area where the second detection unit 213 detected smoke and flames.

[0192] Furthermore, in this embodiment, the detection device 210 displays on the display unit the area in the waste pit 101 where the first detection unit 212 has detected smoke and flames, and the second detection unit 213 has detected smoke and flames, overlapping with each other. For example, a flame detection result display screen 500 having a first flame detection result display area where the overlapping detection area 513 is displayed is displayed on a display unit installed in, for example, the central control room, the crane operation room, and the external control device. Alternatively, a flame detection result display screen 500 having a second flame detection result display area 520 that displays overlapping addresses between the addresses detected by the first detection unit 212 and the addresses detected by the second detection unit 213, enclosed in a rectangular frame, is displayed on a display unit installed in, for example, the central control room, the crane operation room, and the external control device. Therefore, the areas where smoke and flames were detected by the first detection unit 212 and the areas where smoke and flames were detected by the second detection unit 213 can be clearly shown to at least one operator of the central control room, the crane operation room, and the external control device.

[0193] Furthermore, in this embodiment, the detection device 210 notifies the second notification unit of the detection result of smoke or flame generation when the conditions corresponding to each detection mode are met, depending on whether the mode information is one of the first detection mode, the second detection mode, or the third detection mode. Specifically, when the mode information indicates the first detection mode, the detection result of the first detection unit 212 is notified when the first detection unit 212 detects the generation of smoke or flame. When the mode information indicates the second detection mode, the detection result of the second detection unit 213 is notified when the second detection unit 213 detects the generation of smoke or flame. Also, when the mode information indicates the third detection mode, the detection results of the first detection unit 212 and the second detection unit 213 are notified when both the first detection unit 212 and the second detection unit 213 detect the generation of smoke or flame. Therefore, the detection of smoke and flames using a trained model and the detection using an invisible light sensor can be selected by at least one operator, for example, in the central control room, the crane operation room, and the external control device. For example, each operator can use the first detection mode if they want to detect the occurrence of smoke and flames early, the second detection mode if they want to detect the occurrence of smoke and flames with high accuracy (high reliability), and the third detection mode if they want to detect the occurrence of smoke and flames with even higher accuracy (higher reliability).

[0194] Furthermore, in this embodiment, when the detection device 210 detects the occurrence of smoke or flames based on either the learned model or the invisible light sensor, it performs smoke or flame detection processing based on at least one of the remaining sensors. Therefore, the occurrence of smoke or flames in the waste pit 110 can be detected more reliably.

[0195] Furthermore, in this embodiment, the detection device 210 causes the other of the first detection unit 212 and the second detection unit 213 to perform smoke / flame detection processing when one of them detects the occurrence of smoke / flames. For example, when one of the first detection unit 212 and the second detection unit 213 detects the occurrence of smoke / flames, an interrupt processing can be performed on the other unit, which detects the occurrence of smoke / flames at regular intervals, to detect the occurrence of smoke / flames. Therefore, the occurrence of smoke / flames in the waste pit 110 can be detected more reliably.

[0196] Furthermore, in this embodiment, the detection device 210 uses an enlarged version of the captured image data (in-pit image data) corresponding to the area where the second detection unit 213 detected the generation of smoke and flames as estimation image data (in-pit image data), and causes the first detection unit 212 to perform smoke and flame detection processing. Therefore, the generation of smoke and flames in the waste pit 110 can be detected more reliably using the learned model.

[0197] Furthermore, in this embodiment, the detection device 210 causes the third notification unit to notify the first detection result information indicating the detection result of the first detection unit 212 when the first detection unit 212 detects smoke or flames and the reliability information indicating the reliability of the detection result exceeds a predetermined threshold. The detection device 210 lowers the predetermined threshold when the first detection unit 212 does not detect smoke or flames, but the second detection unit 213 detects smoke or flames. Therefore, the threshold for the reliability of the detection result of smoke or flame generation by the first detection unit 212 can be optimized. Consequently, false detections (e.g., non-detection or over-detection) of smoke or flame generation by the first detection unit 212 can be reduced.

[0198] Furthermore, in this embodiment, the detection device 210 controls the water discharge unit to discharge water onto the smoke or flames when both the first detection unit 212 and the second detection unit 213 detect smoke or flames. Therefore, water can be discharged into the waste pit 110 when there is a high probability of smoke or flames being generated.

[0199] Furthermore, in this embodiment, the detection device 210 causes the determination unit to determine whether or not to control the water discharge unit to discharge water onto the smoke and flames, based on the first detection result information indicating the detection result of the first detection unit 212 and the second detection result information indicating the detection result of the second detection unit 213. Therefore, water can be discharged onto the garbage pit 110 according to the result of the determination unit's determination, thus enabling appropriate water discharge onto the garbage pit 110.

[0200] Furthermore, in this embodiment, the detection device 210 transmits first detection result information, which indicates the detection result of the first detection unit 212, and second detection result information, which indicates the detection result of the second detection unit 213, to an external control device 230 installed in a different area from the area where the garbage pit 110 is located. Subsequently, the detection device 210 receives control information from the external control device 230 indicating whether or not to control the water discharge unit to discharge water to the smoke and flames. Therefore, water can be discharged to the garbage pit 110 according to the judgment of the external control device 230. Thus, water can be discharged to the garbage pit 110 appropriately.

[0201] Furthermore, in this embodiment, if the first detection unit 212 or the second detection unit 213 detects smoke or flames after water has been sprayed onto the smoke or flames by the water discharge unit, the detection device 210 transmits first detection result information indicating the detection result of the first detection unit 212 and second detection result information indicating the detection result of the second detection unit 213 to the external control device 230. Therefore, it is possible to inform the operator of the external control device 230 that there is a possibility that the smoke or flames have not been extinguished even after water has been sprayed. Thus, it is possible to prompt the external control device 230 to take appropriate action.

[0202] Furthermore, in this embodiment, the detection device 210 stores in the storage unit 215 the acquired estimation image data (pit image data) in which the first detection unit 212 does not detect the occurrence of smoke or flames, but the second detection unit 213 does detect the occurrence of smoke or flames. Therefore, this estimation pit image data can be used to retrain the trained model. Thus, the detection accuracy of the trained model can be further improved.

[0203] Furthermore, in this embodiment, the detection device 210 uses a model that includes a first trained model, which learns the relationship between captured image data as training data and training data indicating whether or not smoke or flames are included in the subject of the captured image data, and a language model 221. Therefore, the generation of smoke or flames can be detected by taking advantage of the strengths of each trained model.

[0204] Furthermore, in this embodiment, the detection device 210 (first detection unit 212) causes the language model 221 to estimate the occurrence of smoke or flames when the first trained model estimates the occurrence of smoke or flames, or when the second detection unit 213 detects the occurrence of smoke or flames. Therefore, even when the occurrence of smoke or flames is detected based on a trained model other than the language model 221 or an invisible light sensor, the language model 221 can be made to estimate the occurrence of smoke or flames. Thus, it becomes possible to reduce the cost of having the language model 221 detect the occurrence of smoke or flames.

[0205] Furthermore, in this embodiment, the detection device 210 (first detection unit 212) causes the language model 221 to estimate the occurrence of smoke and flames when the confidence level of the first trained model's estimation of smoke and flame occurrence is above a predetermined value and below a predetermined value. Therefore, for example, when the confidence level of the first trained model's estimation of smoke and flame occurrence is neither high nor low (when the first trained model does not infer with confidence), the language model 221 can be made to estimate the occurrence of smoke and flames. Thus, it becomes possible to reduce the cost of having the language model 221 detect the occurrence of smoke and flames.

[0206] Furthermore, in this embodiment, the detection device 210 (first detection unit 212) causes the language model 221 to estimate the occurrence of smoke or flames when it detects smoke or flames based on the first trained model after water is sprayed onto the smoke or flames by the water discharge unit, or when the second detection unit 213 detects smoke or flames. Therefore, even if water is sprayed, the language model 221 can be made to estimate the occurrence of smoke or flames if the occurrence of smoke or flames is detected based on the first trained model or the invisible light sensor. Thus, it becomes possible to reduce the cost of having the language model 221 detect the occurrence of smoke or flames.

[0207] Furthermore, in this embodiment, the language model 221 estimates the generation of smoke and flames, and the type of waste stored in the waste pit 110. Therefore, when the language model 221 is made to estimate the generation of smoke and flames, it can also be made to estimate the type of waste stored in the waste pit 110. Thus, it becomes possible to reduce the cost by having the language model 221 perform both the estimation of smoke and flame generation and the estimation of the type of waste stored in the waste pit 110.

[0208] Furthermore, in this embodiment, the language model 221 estimates the occurrence of smoke and flames based on estimated instruction information transmitted from an external control device 230 installed in a different area from the area where the waste pit 110 is located. Therefore, the language model 221 can be made to estimate the occurrence of smoke and flames when there is an instruction from the external control device 230. Thus, it is possible to reduce the cost of having the language model 221 detect the occurrence of smoke and flames.

[0209] (Other embodiments) The embodiments of this disclosure described above can be implemented by a computer executing a program. Furthermore, a computer-readable recording medium on which the program is stored, and a computer program product such as the program itself, can also be applied as embodiments of this disclosure. Examples of recording media include flexible disks, hard disks, optical disks, magneto-optical disks, CD-ROMs, magnetic tapes, non-volatile memory cards, ROMs, etc. Moreover, the embodiments of this disclosure may be implemented by a PLC (Programmable Logic Controller) or by dedicated hardware such as an ASIC (Application Specific Integrated Circuit). Furthermore, the embodiments of this disclosure described above are merely examples of concrete implementations of this disclosure, and the technical scope of this disclosure should not be interpreted as being limited by them. In other words, this disclosure can be implemented in various ways without departing from its technical concept or its main features.

[0210] Furthermore, the disclosure of the above embodiments is as follows, for example. [Disclosure 1] An acquisition means for acquiring estimated image data showing a photographic image of a waste pit where waste is stored, A first detection means that receives captured image data showing the aforementioned captured image and detects the occurrence of at least one of smoke and flames based on a trained model that estimates the occurrence of at least one of smoke and flames in the waste pit, and the estimation captured image data acquired by the acquisition means, A second detection means that uses an invisible light sensor to detect the generation of at least one of the smoke and flames, A system characterized by comprising [a certain feature]. [Disclosure 2] When at least one of the first detection means and the second detection means detects the occurrence of at least one of the smoke and flame, the first control means causes the first notification unit to report detection result information indicating the respective detection results of the first detection means and the second detection means. The detection system according to disclosure 1, further comprising the above. [Disclosure 3] The first notification unit includes a display unit, The first control means displays on the display unit, in a comparable manner, first area information indicating a first area in the waste pit where the first detection means has detected at least one of the smoke and flames, and second area information indicating a second area in the waste pit where the second detection means has detected at least one of the smoke and flames. The detection system according to disclosure 2, characterized in that [Disclosure 4] The first control means causes the display unit to display, in a discriminable manner, the area in the waste pit where the first area indicated by the first area information and the second area indicated by the second area information overlap with each other. The detection system according to disclosure 3, characterized by the features described herein. [Disclosure 5] A first storage control means that causes a storage unit to store mode information indicating one of the first detection mode, the second detection mode, and the third detection mode, When the mode information stored in the memory unit indicates a first detection mode, the second control means causes the second notification unit to notify the first detection unit of the detection result of the first detection means when the first detection means detects the occurrence of at least one of the smoke and flames; when the mode information stored in the memory unit indicates a second detection mode, the second notification unit is caused to notify the second detection unit of the detection result of the second detection means when the second detection means detects the occurrence of at least one of the smoke and flames; and when the mode information stored in the memory unit indicates a third detection mode, the second control means causes the second notification unit to notify the second detection unit of the detection result of the first detection means and the second detection means, respectively. A detection system according to any one of disclosures 1 to 4, further comprising the above. [Disclosure 6] When the first detection means or the second detection means detects the occurrence of at least one of the smoke and flames based on either the trained model or the invisible light sensor, the third control means causes at least one of the first detection means and the second detection means to perform a detection process for the occurrence of at least one of the smoke and flames based on the remaining at least one of the smoke and flames that has not been detected. A detection system according to any one of disclosures 1 to 5, further comprising the above. [Disclosure 7] A third control means that, when one of the first detection means and the second detection means detects the occurrence of at least one of the smoke and flames, causes the other to perform a detection process for the occurrence of at least one of the smoke and flames. A detection system according to any one of disclosures 1 to 6, further comprising the above. [Disclosure 8] One of the first detection means and the second detection means is the second detection means, The third control means uses an enlarged version of the captured image data corresponding to the area in the waste pit where the second detection means detected the occurrence of at least one of the smoke and flames as the estimation image data, and causes the first detection means to perform the detection process for at least one of the smoke and flames. The detection system according to disclosure 7, characterized by the features described herein. [Disclosure 9] A fourth control means causes the third notification unit to report first detection result information indicating the detection result of the first detection means when the first detection means detects at least one of the smoke and flames, and the reliability information indicating the reliability of the detection result exceeds a predetermined threshold, If the first detection means does not detect at least one of the smoke and flame, and the second detection means detects at least one of the smoke and flame, an adjustment means to lower the predetermined threshold, A detection system according to any one of disclosures 1 to 8, further comprising the above. [Disclosure 10] When both the first detection means and the second detection means detect at least one of the smoke and flames, a fifth control means controls the water discharge unit to discharge water onto at least one of the smoke and flames. A detection system according to any one of disclosures 1 to 9, further comprising the above. [Disclosure 11] A sixth control means causes a determination unit to determine whether or not to control the water discharge unit to discharge water towards at least one of the smoke and flames, based on first detection result information indicating the detection result of the first detection means and second detection result information indicating the detection result of the second detection means. A detection system according to any one of disclosures 1 to 10, further comprising the above. [Disclosure 12] A transmission means for transmitting first detection result information indicating the detection result of the first detection means and second detection result information indicating the detection result of the second detection means to an external control device installed in a different area from the area where the waste pit is located, After the transmission process of the transmission means, a receiving means receives control information from the external control device indicating whether or not to control the water discharge unit to discharge water onto at least one of the smoke and flames, A detection system according to any one of disclosures 1 to 11, further comprising the above. [Disclosure 13] The transmitting means transmits the first detection result information and the second detection result information to the external control device when the first detection means or the second detection means detects at least one of the smoke and flames after the water discharge unit has discharged water onto at least one of the smoke and flames. The detection system according to disclosure 12, characterized in that [Disclosure 14] A second storage control means causes the storage unit to store in the storage unit the estimated photographic image data acquired by the acquisition means, corresponding to cases where the first detection means does not detect the occurrence of at least one of the smoke and flames, and the second detection means detects the occurrence of at least one of the smoke and flames. A detection system according to any one of disclosures 1 to 13, further comprising the above. [Disclosure 15] The invisible light sensor includes at least one of an infrared sensor and an ultraviolet sensor. A detection system according to any one of disclosures 1 to 14, characterized in that [Disclosure 16] The aforementioned trained model is A first trained model that has learned the relationship between the captured image data as training data and the training data indicating whether or not the subject of the captured image data contains at least one of the smoke and flames, Language models and, including, A detection system according to any one of disclosures 1 to 15, characterized in that [Disclosure 17] The first detection means causes the language model to estimate the occurrence of at least one of the smoke and flames when the first trained model estimates the occurrence of at least one of the smoke and flames, or when the second detection means detects the occurrence of at least one of the smoke and flames. The detection system according to disclosure 16, characterized in that [Disclosure 18] The first detection means causes the language model to estimate the occurrence of at least one of the smoke and flames when the confidence level of the estimation of the occurrence of at least one of the smoke and flames by the first trained model is above a predetermined value and below a predetermined value. The detection system according to disclosure 17, characterized in that [Disclosure 19] The first detection means, after spraying water onto at least one of the smoke and flames by the water discharge unit, causes the language model to estimate the occurrence of at least one of the smoke and flames when it detects at least one of the smoke and flames based on the first trained model, or when the second detection means detects at least one of the smoke and flames. A detection system according to any one of disclosures 16 to 18, characterized in that [Disclosure 20] The language model estimates the generation of at least one of the smoke and flames, and the type of waste stored in the waste pit. A detection system according to any one of disclosures 16 to 19, characterized in that [Disclosure 21] The language model estimates the occurrence of at least one of the smoke and flames based on estimated instruction information transmitted from an external control device installed in a different area from the area where the waste pit is located. A detection system according to any one of disclosures 16 to 20, characterized in that [Disclosure 22] An acquisition step to obtain estimated image data showing a photographic image of a waste pit where waste is stored, A first detection step is performed to detect the occurrence of at least one of the smoke and flames based on a trained model that receives the captured image data and estimates the occurrence of at least one of the smoke and flames in the waste pit, and the estimation captured image data acquired in the acquisition step. A second detection step involves using an invisible light sensor to detect the generation of at least one of the smoke and flames, A detection method characterized by comprising the following features. [Disclosure 23] A program for causing a computer to function as each means of the detection system described in any one of disclosures 1 to 21. [Explanation of symbols]

[0211] 110 Waste Storage Pit 120a~120b Crane 121a~121b Traveling device 122a~122b Traverse device 123a~123b Traverse rails 124a~124b Buckets 130 Running Rails 140a~140b Imaging equipment 150a~150b Water spray device 151a~151b Pump 152a~152b Solenoid valve 160 UV Sensor 170 Infrared Sensor 180 Garbage Disposal Gate 190 waste 210 Detection device 211 Acquisition Department 212 First detection unit 213 Second detection unit 214 Control Unit 215 Storage section 216 Communications Department 217 Input Reception Section 220 Language Model Servers 221 Language Models 230 External control device 240 Processing Units 250, 260 networks 400 detection prompts 401-405 Prompts 1-5 410 Prompt for requesting water discharge determination 411-415 Prompts 1-5 500 Flame detection result display screen 510 First flame detection result display area 511 Pit simulation image 512a~512c Flame detection area 513 Duplicate detection area 514a~514b Crane position information 520 Second flame detection result display area 530 Crane position display area

Claims

1. An acquisition means for acquiring estimated image data showing a photographic image of a waste pit where waste is stored, A first detection means that receives the captured image data and a trained model that estimates the occurrence of at least one of smoke and flames in the waste pit, and estimates the captured image data acquired by the acquisition means, A second detection means that uses an invisible light sensor to detect the generation of at least one of the smoke and flames, A system characterized by comprising [a certain feature].

2. When at least one of the first detection means and the second detection means detects the generation of at least one of the smoke and flame, the first control means causes the first notification unit to report detection result information indicating the respective detection results of the first detection means and the second detection means. The detection system according to claim 1, further comprising the following:

3. The first notification unit includes a display unit, The first control means displays on the display unit, in a comparable manner, first area information indicating a first area in the waste pit where the first detection means has detected at least one of the smoke and flames, and second area information indicating a second area in the waste pit where the second detection means has detected at least one of the smoke and flames. The detection system according to feature 2.

4. The first control means causes the display unit to display, in a discriminable manner, the area in the waste pit where the first area indicated by the first area information and the second area indicated by the second area information overlap with each other. The detection system according to feature 3.

5. A first storage control means that causes a storage unit to store mode information indicating one of the first detection mode, the second detection mode, and the third detection mode, When the mode information stored in the memory unit indicates a first detection mode, the second control means causes the second notification unit to notify the first detection unit of the detection result of the first detection means when the first detection means detects the occurrence of at least one of the smoke and flames; when the mode information stored in the memory unit indicates a second detection mode, the second notification unit is caused to notify the second detection unit of the detection result of the second detection means when the second detection means detects the occurrence of at least one of the smoke and flames; and when the mode information stored in the memory unit indicates a third detection mode, the second control means causes the second notification unit to notify the second detection unit of the detection result of the first detection means and the second detection means, respectively. The detection system according to any one of claims 1 to 4, further comprising the above.

6. When the first detection means or the second detection means detects the occurrence of at least one of the smoke and flames based on either the trained model or the invisible light sensor, the third control means causes at least one of the first detection means and the second detection means to perform a detection process for the occurrence of at least one of the smoke and flames based on the remaining at least one of the smoke and flames that has not been detected. The detection system according to any one of claims 1 to 4, further comprising the above.

7. A third control means that, when one of the first detection means and the second detection means detects the occurrence of at least one of the smoke and flames, causes the other to perform a detection process for the occurrence of at least one of the smoke and flames. The detection system according to any one of claims 1 to 4, further comprising the above.

8. One of the first detection means and the second detection means is the second detection means, The third control means uses an enlarged version of the captured image data corresponding to the area in the waste pit where the second detection means detected the occurrence of at least one of the smoke and flames as the estimation image data, and causes the first detection means to perform the detection process for at least one of the smoke and flames. The detection system according to feature 7.

9. A fourth control means causes a third notification unit to report first detection result information indicating the detection result of the first detection means when the first detection means detects at least one of the smoke and flames, and the reliability information indicating the reliability of the detection result exceeds a predetermined threshold, If the first detection means does not detect at least one of the smoke and flame, and the second detection means detects at least one of the smoke and flame, an adjustment means to lower the predetermined threshold, The detection system according to any one of claims 1 to 4, further comprising the above.

10. A fifth control means controls the water discharge unit to discharge water onto at least one of the smoke and flames when both the first detection means and the second detection means detect at least one of the smoke and flames. The detection system according to any one of claims 1 to 4, further comprising the above.

11. A sixth control means causes a determination unit to determine whether or not to control the water discharge unit to discharge water towards at least one of the smoke and flames, based on first detection result information indicating the detection result of the first detection means and second detection result information indicating the detection result of the second detection means. The detection system according to any one of claims 1 to 4, further comprising the above.

12. A transmission means for transmitting first detection result information indicating the detection result of the first detection means and second detection result information indicating the detection result of the second detection means to an external control device installed in a different area from the area where the waste pit is located, After the transmission process of the transmission means, a receiving means receives control information from the external control device indicating whether or not to control the water discharge unit to discharge water onto at least one of the smoke and flames, The detection system according to any one of claims 1 to 4, further comprising the above.

13. The transmitting means transmits the first detection result information and the second detection result information to the external control device when the first detection means or the second detection means detects at least one of the smoke and flames after the water discharge unit has discharged water onto at least one of the smoke and flames. The detection system according to feature 12.

14. A second storage control means causes the storage unit to store in the storage unit the estimation image data acquired by the acquisition means, corresponding to cases where the first detection means does not detect the occurrence of at least one of the smoke and flames, and the second detection means detects the occurrence of at least one of the smoke and flames. The detection system according to any one of claims 1 to 4, further comprising the above.

15. The invisible light sensor includes at least one of an infrared sensor and an ultraviolet sensor. The detection system according to any one of claims 1 to 4.

16. The aforementioned trained model is A first trained model that has learned the relationship between the captured image data as training data and the training data indicating whether or not the subject of the captured image data contains at least one of the smoke and flames, Language models and, including, The detection system according to any one of claims 1 to 4.

17. The first detection means causes the language model to estimate the occurrence of at least one of the smoke and flames when the first trained model estimates the occurrence of at least one of the smoke and flames, or when the second detection means detects the occurrence of at least one of the smoke and flames. The detection system according to claim 16.

18. The first detection means causes the language model to estimate the occurrence of at least one of the smoke and flames when the confidence level of the estimation of the occurrence of at least one of the smoke and flames by the first trained model is above a predetermined value and below a predetermined value. The detection system according to feature 17.

19. The first detection means, after spraying water onto at least one of the smoke and flames by the water discharge unit, causes the language model to estimate the occurrence of at least one of the smoke and flames when it detects at least one of the smoke and flames based on the first trained model, or when the second detection means detects at least one of the smoke and flames. The detection system according to claim 16.

20. The language model estimates the generation of at least one of the smoke and flames, and the type of waste stored in the waste pit. The detection system according to claim 16.

21. The language model estimates the occurrence of at least one of the smoke and flames based on estimated instruction information transmitted from an external control device installed in a different area from the area where the waste pit is located. The detection system according to claim 16.

22. An acquisition step to obtain estimated image data showing a photographic image of a waste pit where waste is stored, A first detection step is performed to detect the occurrence of at least one of the smoke and flames based on a trained model that receives the captured image data and estimates the occurrence of at least one of the smoke and flames in the waste pit, and the estimation captured image data acquired in the acquisition step. A second detection step involves using an invisible light sensor to detect the generation of at least one of the smoke and flames, A detection method characterized by comprising the following features.

23. A program for causing a computer to function as each means of the detection system according to any one of claims 1 to 4.

Images