Information processing device, information processing method and program

Abstract

To grasp properties of waste at a step prior to being loaded into an incinerator, to predict changes in a state inside the incinerator, and to enable stable operation of a waste incineration plant.SOLUTION: A control unit of an information processing device is configured to: acquire image information from a photographing unit which captures waste when being loaded into a loading unit configured to load and supply waste to an incinerator; extract a still image of the waste captured in the image information; input the still image as an input parameter to an image analysis model; and classify a plurality of types of waste included in the still image by type and output the result as output parameters of the image analysis model. The image analysis model is a learning model generated by machine learning, using still images extracted from video information capturing a state of waste being loaded into the loading unit as learning input parameters, and images of the waste in the still images classified by type as learning output parameters.SELECTED DRAWING: Figure 10

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] Traditionally, it has been important to understand the contents of waste in the operation of waste incineration plants. However, currently, most waste incineration plants do not detect the contents of waste before combustion, but instead use a method of estimating the contents based on measurements inside the incineration furnace. In methods that estimate the contents based on measurements inside the incineration furnace, even if the contents of the waste change, information on the change in the contents can only be obtained after combustion. Therefore, adjustments to combustion conditions during waste incineration operations are made after changes in the operating conditions have appeared.

[0003] If it were possible to grasp the contents of waste before combustion, it would be possible to adjust the combustion conditions in advance according to the properties of the waste contents. If changes in the operating state could be suppressed, it would be possible to stabilize the operation of waste incineration plants. For this reason, methods for grasping the operating state before combustion have been proposed.

[0004] Patent Document 1 describes an apparatus that includes a training data generation unit that generates training data associated with images of waste stored in a waste pit, a model construction unit that constructs a model by learning using the training data, and an estimation unit that inputs new image data of waste stored in the waste pit into the model and obtains a value representing the quality of the waste corresponding to the new image. That is, Patent Document 1 discloses a configuration that analyzes information about waste in a waste pit using image analysis to determine the quality of the waste in order to understand the contents of the waste in advance and improve the operation of a waste incineration plant.

[0005] Patent Document 2 describes a prediction device for an incineration facility where materials to be incinerated are fed into a hopper and sent to an incinerator at a specified speed, the prediction device comprising: a movement data generation unit that generates movement data showing the movement state of the materials to be incinerated from multiple time-series images taken from above the hopper, and a property prediction unit that predicts the properties of the materials to be incinerated based on the generated movement data and the speed. That is, Patent Document 2 discloses a method for measuring the amount of waste in a waste hopper and estimating the calories of the waste from the speed at which the waste falls in the hopper chute. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2019-207099 [Patent Document 2] Japanese Patent Publication No. 2022-168717 Summary of the Invention [Problem to be solved by the invention]

[0007] However, since the camera used in the technology described in Patent Document 1 is installed in the waste storage pit, the distance between the camera and the waste is large, making it difficult to grasp the contents of the waste in detail. Furthermore, the waste properties that can be grasped by the technology described in Patent Document 1 are highly dependent on the properties of the waste located on the top surface of the waste stored in the storage pit, so there is a high possibility that there will be a large discrepancy between the properties of the waste actually grasped by the crane bucket.

[0008] Furthermore, the technology described in Patent Document 2 has the problem that the prediction formula varies greatly depending on the shape of the input hopper. Furthermore, during operation of a waste incineration plant, the height of the waste in the hopper is not always at a height that allows the technology described in Patent Document 2 to be adopted. For example, if the height of the top surface of the waste is so low that it cannot be seen by an imaging device, it becomes extremely difficult to adopt the technology described in Patent Document 2.

[0009] In other words, while the above-mentioned conventional technologies were able to ascertain the properties of the waste inside the storage pit or the input hopper to a certain extent, they lacked accuracy in determining the properties of the waste. Therefore, there has been a demand for the development of a technology that can predict fluctuations in the combustion conditions inside the incinerator and enable stable operation of waste incineration plants by more accurately ascertaining the properties of the waste before it is input into the incinerator, without relying on the height of the waste stored in the waste storage pit or the condition of the outermost surface.

[0010] The present invention has been made in consideration of the above circumstances, and its purpose is to provide an information processing device, an information processing method, and a program that can grasp the properties of waste before it is put into the incinerator, predict changes in the situation inside the incinerator, and enable stable operation of a waste incineration plant. [Means for solving the problem]

[0011] In order to solve the above-mentioned problems and achieve the objective, an information processing device according to one embodiment of the present invention includes a control unit having hardware, and the control unit acquires image information from an imaging unit having at least one imaging camera capable of capturing images of the waste as it is being fed into an input unit configured to allow waste to be fed into the input unit and to supply it to an incinerator, extracts a still image of the waste included in the image information, inputs the still image as an input parameter to an image analysis model, and classifies multiple types of waste included in the waste in the still image by type and outputs them as output parameters of the image analysis model, and the image analysis model is a learning model generated by machine learning using still images extracted from video information capturing the state of waste being fed into the input unit as learning input parameters and images classified by type for the waste captured in the still image as learning output parameters.

[0012] In one aspect of the information processing device of the present invention, in the above invention, the imaging unit is located at a height that is in the range of -45 degrees or more and 45 degrees or less from the horizontal relative to the height at which the gripping unit, which can grip and release the waste, puts the waste into the input unit.

[0013] In one embodiment of the information processing device of the present invention, in the above invention, the imaging unit is equipped with a plurality of imaging cameras, and at least two of the plurality of imaging cameras are arranged so that the angle at which the waste is imaged is within a range of greater than 0° and less than 180° along a horizontal plane, centered on the position at which the gripping unit capable of grasping and releasing the waste releases the waste.

[0014] In one embodiment of the information processing device of the present invention, in the above invention, the imaging unit is equipped with a plurality of imaging cameras, and at least two of the plurality of imaging cameras are arranged so that the angle at which the waste is imaged is within a range of 90° or more and 180° or less along a horizontal plane, centered on the position at which the gripping unit capable of grasping and releasing the waste releases the waste.

[0015] In the information processing device according to one aspect of the present invention, in the above invention, the image analysis model is a learning model generated by an algorithm using semantic segmentation.

[0016] In one aspect of the present invention, an information processing device classifies multiple types of waste contained in the waste in the still image by type as output parameters of the image analysis model, color-codes each classified type to generate image analysis results, and generates and outputs a processed image in which the image analysis results are superimposed on the still image.

[0017] An information processing method according to one embodiment of the present invention is an information processing method executed by a control unit having hardware, which acquires image information from an imaging unit having at least one imaging camera capable of capturing images of waste as it is being fed into an input unit configured to allow waste to be fed into the input unit and to supply the waste to an incinerator, stores the image information in a memory unit, extracts a still image of the waste contained in the image information read from the memory unit, inputs the still image as an input parameter into an image analysis model, and classifies multiple types of waste contained in the waste in the still image by type and outputs them as output parameters of the image analysis model, and the image analysis model is a learning model generated by machine learning using still images extracted from video information capturing the state of waste being fed into the input unit as learning input parameters and images classified by type for the waste captured in the still image as learning output parameters.

[0018] In one aspect of the information processing method of the present invention, in the above invention, multiple types of waste contained in the waste in the still image are classified by type as output parameters of the image analysis model, each classified type is color-coded to generate an image analysis result, and a processed image in which the image analysis result is superimposed on the still image is generated and output.

[0019] A program according to one embodiment of the present invention is a program to be executed by a control unit having hardware, and causes the control unit to acquire image information from an imaging unit having at least one imaging camera capable of capturing images of the waste as it is being loaded into an input unit configured to allow waste to be loaded and to supply it to an incinerator, store the image information in a memory unit, extract a still image of the waste included in the image information read from the memory unit, input the still image as an input parameter to an image analysis model, and classify multiple types of waste included in the waste in the still image by type and output the classified images as output parameters of the image analysis model, wherein the image analysis model is a learning model that has been generated in advance by machine learning, using still images extracted from video information capturing the state of waste being loaded into the input unit as learning input parameters, and images classified by type for the waste captured in the still image as learning output parameters. [Effects of the Invention]

[0020] The information processing device, information processing method, and program according to the present invention make it possible to grasp the characteristics of waste before it is put into the incinerator and predict changes in the conditions inside the incinerator, thereby enabling stable operation of waste incineration plants. [Brief explanation of the drawings]

[0021] [Figure 1] FIG. 1 is a block diagram showing a waste incineration plant according to one embodiment of the present invention. [Figure 2] FIG. 2 is a side view illustrating a storage pit, a crane, and a charging hopper according to an embodiment of the present invention. [Figure 3] FIG. 3 is a side view for explaining the storage pit, the crane, the input hopper, and the imaging unit according to one embodiment of the present invention. [Figure 4] FIG. 4 is a top view for explaining the installation position of the imaging unit installed in the input hopper according to one embodiment of the present invention. [Figure 5]FIG. 5 is a block diagram showing an information processing device according to an embodiment of the present invention. [Figure 6] FIG. 6 is a flowchart illustrating a method for generating an image analysis model executed by an information processing device according to an embodiment of the present invention. [Figure 7] FIG. 7 is a diagram showing a state immediately before waste is held in a bucket and thrown into a throwing hopper in a waste incineration plant according to an embodiment of the present invention. [Figure 8] FIG. 8 is a diagram showing a state in which waste is being fed into a feeding hopper in a waste incineration plant according to an embodiment of the present invention. [Figure 9] FIG. 9 is a diagram showing a state in which wastes input into an input hopper in a waste incineration plant according to an embodiment of the present invention are separated into a plurality of types. [Figure 10] FIG. 10 is a flowchart for explaining a method for determining the properties of waste executed by an information processing device according to an embodiment of the present invention. [Figure 11] FIG. 11 is a graph showing the lower heating value (Hu value) and the plastic waste ratio over time in a waste incineration plant controlled by an information processing device according to one embodiment of the present invention. [Figure 12] FIG. 12 is a top view for explaining the installation positions of two imaging units installed in the input hopper according to one embodiment of the present invention. [Figure 13] FIG. 13 is a block diagram showing an information processing device according to a second modified example of one embodiment of the present invention. [Figure 14] FIG. 14 is a flowchart for explaining the learning method in the waste property determination method according to the second modified example of one embodiment of the present invention. [Figure 15] FIG. 15 is a flowchart for explaining the analysis method in the waste property determination method according to the second modified example of one embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0022] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In all the drawings of the embodiment below, the same or corresponding parts are designated by the same reference numerals. Furthermore, the present invention is not limited to the embodiment described below.

[0023] First, in describing an information processing device according to an embodiment of the present invention, the inventors' extensive research will be described. Specifically, the inventors conducted various studies on the above-mentioned conventional technologies to determine the properties of waste before incineration, and discovered the following problem. Specifically, the imaging device employed in Patent Document 1 is installed in the storage pit, which increases the distance between the imaging device and the waste, making it difficult to obtain a detailed understanding of the contents of the waste. Furthermore, the properties of the waste that can be determined using the method described in Patent Document 1 depend heavily on the properties of the waste piled at the surface of the storage pit. Therefore, there is a significant discrepancy between the properties of the waste actually grasped by the gripper.

[0024] Furthermore, the technology proposed in Patent Document 2 determines the moisture content and properties of waste accumulated at the bottom of a hopper. However, in the technology described in Patent Document 2, the prediction formula varies significantly depending on the shape of the input hopper. Furthermore, in the technology described in Patent Document 2, it is necessary to maintain the waste level at a predetermined height or higher, but it is difficult to continuously maintain the height of the waste in the input hopper (waste level) at or above this predetermined height during operation of the waste incineration plant. For example, if the waste level in the input hopper is so low that it cannot be captured by an imaging device, it is difficult to employ the technology described in Patent Document 2.

[0025] Therefore, the present inventors have studied a method that can solve the problems they have encountered and that allows them to grasp the properties of waste in advance. Through their studies, the present inventors have found that, while imaging waste in a storage pit reveals only the surface condition of the waste, imaging waste near a loading hopper allows for closer imaging of the waste. Images of waste near the loading hopper are clearer than images of waste in a storage pit because the waste is closer to the imaging device. Furthermore, the inventors have conceived that, in the case of images of waste near the loading hopper, the waste breaks up when released from a bucket attached to a gripping unit, making it possible to grasp the contents of the waste held in the bucket in a decomposed state.

[0026] Based on the above considerations, the present inventor has devised a method for grasping the properties of waste by installing an imaging device such as a visible light camera above a hopper into which the waste is dropped and capturing images of the waste being moved by a crane, and by using images obtained by this imaging device. Note that the imaging device is not limited to a visible light camera, and a camera capable of capturing images of light (electromagnetic waves) of a specific wavelength, such as an infrared camera, can also be used. Furthermore, the present inventor has devised a method for improving the accuracy of determining the properties of waste by capturing images of the broken-up waste at the moment when it is released from the crane or immediately thereafter, and identifying the type of each piece of waste.

[0027] Here, the imaging device is preferably installed at a position where it can observe the falling of waste when it is released, and at least one visible light camera is used as the imaging device. In order to understand the properties of the waste, the inventor has further devised and studied a method of applying image analysis technology to images of the waste captured by the imaging device. According to the inventor's knowledge, it is desirable to adopt image classification technology using semantic segmentation technology or a convolutional neural network (CNN) as the image analysis technology.

[0028] Furthermore, according to the inventor's findings, semantic segmentation technology is a deep learning algorithm that associates a label or category with every pixel in a given image. Semantic segmentation technology is often used to recognize groups of pixels that form particularly distinctive categories, and is capable of dividing an image of waste into multiple regions at the pixel level. This enables semantic segmentation technology to clearly detect irregularly shaped objects, such as scattered pieces of waste. Therefore, semantic segmentation technology is preferable in that it can perform classification with higher accuracy than other object detection methods by assigning labels to the pixels contained in the image.

[0029] Furthermore, the objects to be classified using image analysis techniques such as semantic segmentation include classification of waste to be incinerated obtained when waste grasped by the gripper is dumped into a feeding hopper, such as plastic waste, plant waste, and sludge waste. This makes it possible to utilize a wider range of information than when simply deriving the descent speed in the feeding hopper, as proposed in the prior art.

[0030] From the above, the present inventor has devised a method for deriving the properties (properties) of waste by classifying the properties of waste into multiple types from images of waste dumped into a feeding hopper and statistically analyzing the information obtained from the images using a method such as regression analysis. Furthermore, according to the knowledge gained by the present inventor from the above-mentioned study, a method for improving accuracy is also possible to analyze the properties of waste by further utilizing sensor information obtained by sensing using a sensor unit, such as the weight of the waste, and adopting a method such as machine learning.

[0031] Based on the above considerations, the inventors have specifically concluded that an imaging device installed in the input hopper is closer to the waste being input than an imaging device installed in the storage pit, thereby enabling the acquisition of more detailed images. Furthermore, when waste is input from a crane into the hopper, it falls in pieces, allowing the entire waste to be observed. By applying image analysis technology (segmentation) to these images, it becomes possible to quantitatively calculate the volume of the entire input waste and the volume of each type of waste. Further, the inventors have concluded that it is preferable to install the imaging device in a location that allows for close imaging of the state in which the waste held by the gripper is released into pieces. Furthermore, the inventors have specifically concluded that the imaging device should be installed at a position on the inner surface of the input hopper at a predetermined angle relative to the release position. The embodiment described below was devised based on the above-mentioned intensive considerations by the inventors.

[0032] (Waste incineration plant) Fig. 1 shows a waste incineration plant to which an information processing device according to one embodiment of the present invention is applied. As shown in Fig. 1, the waste incineration plant 1 includes an information processing device 10, a waste storage facility 20, and a waste incineration facility 30, which are capable of communicating with each other via a network 2. The waste treatment facility 3 includes at least the waste storage facility 20 and the waste incineration facility 30. The information processing device 10 may be installed outside the waste treatment facility 3 and be capable of communicating with the waste treatment facility 3 via the network 2, or may be part of the waste treatment facility 3. The information processing device 10 may also be installed inside the waste storage facility 20, and the installation location is not limited.

[0033] The network 2 is configured by appropriately combining wired communication and wireless communication, and is made up of communication networks such as the Internet network and mobile phone network. The network 2 is made up of one or a combination of, for example, dedicated lines, public communication networks such as the Internet, for example, LANs (Local Area Networks), WANs (Wide Area Networks), telephone communication networks such as mobile phones, public lines, VPNs (Virtual Private Networks), etc. The information processing device 10, waste storage facility 20, and waste incineration facility 30 are connected via the network 2.

[0034] (Waste incineration facility) The waste incineration facility 30, which serves as a waste incineration unit, includes a combustion control device (ACC) 31, a sensor unit 32, and an incinerator 33. The combustion control device 31 controls the combustion air volume, cooling air volume, waste feeder feed speed, and grate feed speed as manipulated variables of each control element based on predetermined manipulated variable reference values. The incinerator 33, which serves as a waste incinerator, includes a furnace where waste 28 is burned, a waste inlet through which waste 28 is introduced, and a boiler (none of which are shown). The sensor unit 32 includes, for example, thermometers and pressure gauges installed in various locations. Various physical quantities, such as pressure and speed, measured by the sensor unit 32 and related conditions within the incinerator 33 and in facilities related to the incinerator 33, specifically, in power generation facilities, are output as sensor information from the sensor unit 32. The sensor information output from the sensor unit 32 is supplied as parameters to the combustion control device 31. The combustion control device 31 controls combustion in the incinerator 33 based on the input parameters.

[0035] (Waste storage facility) The waste storage facility 20 serving as a waste storage unit includes a control unit 21, a communication unit 22, an imaging unit 23, a sensor unit 24, a gripper 25, and a storage pit 26. The storage pit 26 is provided with a platform 27 onto which a refuse collection truck (not shown) or a general vehicle can carry in waste. The storage pit 26 also has a movably mounted gripper 25 above it, and a sensor unit 24. The sensor unit 24 and the gripper 25 are controlled by the control unit 21 based on a control signal transmitted from the control unit 11 of the information processing device 10. The imaging unit 23 is provided near a feed hopper 331 serving as a feed unit into which the waste 28 is fed when transporting the waste 28 from the storage pit 26 to the incinerator 33. The imaging unit 23 is controlled by the control unit 21 based on a control signal transmitted from the control unit 11. The control unit 11 of the information processing device 10 may also directly control the imaging unit 23, the sensor unit 24, and the gripper 25.

[0036] Specifically, control unit 21 includes a processor having hardware such as a CPU (Central Processing Unit), a DSP (Digital Signal Processor), and an FPGA (Field-Programmable Gate Array), and a main storage unit such as a RAM (Random Access Memory) and a ROM (Read Only Memory) (none of which are shown). Control unit 21 controls imaging unit 23, sensor unit 24, and grip unit 25 based on control signals input from information processing device 10 via communication units 13 and 22 in accordance with various programs stored in the main storage unit such as RAM and ROM.

[0037] The communication unit 22 is, for example, a LAN interface board, a wired communication circuit for wired communication, or a wireless communication circuit for wireless communication. The LAN interface board, the wired communication circuit, or the wireless communication circuit is connected to the network 2. The communication unit 22, which functions as a transmitting unit and a receiving unit, is connected to the network 2 and communicates with the information processing device 10.

[0038] The imaging unit 23 includes at least one imaging camera 231. The imaging unit 23 may include multiple cameras, for example, two imaging cameras 231 and 232. The imaging cameras 231 and 232 may be configured to include, for example, CCD cameras or CMOS cameras, but are not limited to these. The imaging cameras 231 and 232 are configured to be able to distinguish the color tone of the waste 28, which changes depending on the type of waste 28. The imaging unit 23 is configured to be able to capture images of the waste 28 gripped by the bucket 253 of the gripper 25 and the waste 28 being released from the bucket 253 using the imaging camera 231 (232). The imaging unit 23 captures images of the waste 28 gripped by the gripper 25 and the state of the waste 28 being released, and transmits the captured image data as imaging information to the information processing device 10 via the communication unit 22.

[0039] The sensor unit 24 is configured to include multiple distance measurement sensors, for example, LiDAR (Laser Imaging Detection and Ranging). Note that the sensor unit 24 may be configured with, in addition to distance measurement sensors, for example, a laser sensor, an infrared sensor, or a sensor that combines these sensors. The sensor unit 24 transmits distance information such as the measured distance to the information processing device 10 via the communication unit 22. The storage pit 26 is a pit that can temporarily store waste 28.

[0040] The gripping unit 25 grips and moves the waste 28 stored in the storage pit 26. FIG. 2 is a side view illustrating the storage pit 26, crane 252, and input hopper 331 according to this embodiment. As shown in FIG. 2, the gripping unit 25 includes a crane 252 and a bucket 253. The crane 252 serving as a moving unit is configured to be movable by connecting the bucket 253. The crane 252 and the bucket 253 are configured to be able to reciprocate between above the storage pit 26 and above the input hopper 331. The bucket 253 serving as an opening / closing unit is configured to be able to move up and down, and the bucket 253 can descend to grip the waste 28.

[0041] In the storage pit 26, the bucket 253 of the gripper 25 rises while gripping the waste 28, and then moves horizontally to above the input hopper 331 and dumps the waste 28. This transports the waste 28 from the storage pit 26 to the input hopper 331. That is, the waste 28 in the storage pit 26 is gripped by the gripper 25 and supplied to the incinerator 33 of the waste incineration facility 30 through the input hopper 331, where it is incinerated.

[0042] Here, parameters related to the work operation of the crane 252, such as the current position of the crane 252 in two dimensions (x, y) in a horizontal plane relative to the storage pit 26, the movement path, the movement range, and the movement speed, are sequentially transmitted from the gripper 25 to the information processing device 10 via the control unit 21 and the communication unit 22. Similarly, parameters related to the work operation of the bucket 253, such as the current position of the bucket 253 in three dimensions (x, y, z), the open time, the close time, the number of times it is opened and closed, and the amount of opening, are sequentially transmitted from the gripper 25 to the information processing device 10 via the control unit 21 and the communication unit 22.

[0043] 3 and 4 are side and top views, respectively, illustrating the storage pit 26, crane 252, input hopper 331, and imaging unit 23 according to this embodiment. As shown in FIGS. 3 and 4, the imaging unit 23 is provided on the wall surface above the input hopper 331, at a position where it can capture an image at a predetermined angle θ from the horizontal direction relative to the position where the gripper 25 releases the waste 28. The predetermined angle θ is set, for example, within a range of −45 degrees to 45 degrees (−45°≦θ≦45°). In the example shown in FIG. 4, the imaging unit 23 is provided at one of the four upper corners of the wall surface of the input hopper 331, but it may also be provided at another corner or on a side, and is not limited thereto.

[0044] (Information processing device) Fig. 5 shows details of the information processing device 10 in Fig. 1. The information processing device 10 shown in Fig. 5 includes a control unit 11, a storage unit 12, a communication unit 13, and an input / output unit 14. The control unit 11 and the communication unit 13 are physically similar to the control unit 21 and the communication unit 22 described above, respectively.

[0045] The input / output unit 14 may be configured, for example, with a touch panel display or a speaker / microphone. The input / output unit 14 as an input means includes an interface that receives various information transmitted via the communication unit 22 from the imaging unit 23 or the sensor unit 24 installed in the waste storage facility 20 and outputs the information to the control unit 11. Information may be transmitted from the imaging unit 23 or the sensor unit 24 to the input / output unit 14 via wired or wireless communication. The input / output unit 14 also includes a user interface, such as a keyboard, input buttons, a lever, a touch panel for manual input superimposed on a display such as an LCD, or a microphone for voice recognition. A worker or the like can input predetermined information to the control unit 11 by operating the input / output unit 14. The input / output unit 14 as an output means, under the control of the control unit 11, displays an image of the bucket 253 of the gripper 25 moving into the waste storage facility 20's input hopper 331 on a display monitor, displays text and figures on a touch panel display screen, and outputs sound from a speaker. That is, the input / output unit 14 is configured to be able to notify predetermined information to the outside. The input unit and the output unit in the input / output unit 14 may be configured as separate units.

[0046] The storage unit 12 is configured with a storage medium selected from volatile memory such as RAM, non-volatile memory such as ROM, erasable programmable ROM (EPROM), hard disk drive (HDD), and removable media. The removable media is, for example, a universal serial bus (USB) memory or a disc storage medium such as a compact disc (CD), a digital versatile disc (DVD), or a Blu-ray (registered trademark) disc (BD). Alternatively, the storage unit 12 may be configured with a computer-readable storage medium such as an externally attachable memory card, or may be configured with cloud storage capable of storing data via the network 2.

[0047] The storage unit 12 can store an operating system (OS), various programs, various tables, various databases, and the like for executing the operations of the information processing device 10. These various programs can also be recorded on computer-readable recording media such as a hard disk, flash memory, CD-ROM, DVD-ROM, or flexible disk and widely distributed. The control unit 11 loads the programs stored in the storage unit 12 into a working area of the main storage unit, executes them, and controls each component through the execution of the programs, thereby realizing functions that meet predetermined purposes.

[0048] In this embodiment, the functions of an image processing unit 111, a type determination unit 112, a waste information generation unit 113, an annotation unit 114, a crane control unit 115, and a learning unit 116 are performed by executing the various programs loaded by the control unit 11. The various programs also include programs that realize artificial intelligence and learned models that can realize the processing according to this embodiment.

[0049] The image processing unit 111 is capable of performing predetermined image processing on the acquired image information. The image processing unit 111 is configured to be capable of performing image processing to cut out and extract still images from video data input from the imaging unit 23. The image processing unit 111 also performs image processing on the acquired still images, such as assigning a preset color to predetermined regions of the still images, for example, regions where a predetermined type of waste 28 exists for each type of waste 28.

[0050] The type determination unit 112 determines the type (also referred to as classification) of each waste contained in the waste 28 when it is thrown into the throw-in hopper 331. The type determination unit 112 may determine the state of the waste 28 based on imaging information supplied from the imaging cameras 231, 232.

[0051] The waste information generation unit 113 generates information that distinguishes the waste 28 included in the image information 122 into predetermined classified areas according to the type of waste 28 determined by the type determination unit 112. That is, as a result of the type determination unit 112 determining the type of each waste 28 included in the waste 28, the waste information generation unit 113 associates the type of waste 28 included in the image information 122 with the area into which the waste 28 is classified, thereby generating waste information 121. The waste information 121 includes numerical information as output parameters output from an image analysis model 123, which will be described later. Note that the waste information generation unit 113 is capable of generating waste information 121 discontinuously and intermittently for each still image included in the acquired video.

[0052] The annotation unit 114 performs labeling on still images extracted by the image processing unit 111 from the video acquired from the imaging unit 23, based on data input by the worker using the input / output unit 14. Specifically, the worker sets a label based on the type of waste for the still image (also called an image patch) using the input / output unit 14. The label set for the still image is associated with the coordinates of the pixel in the still image and stored as waste information 121 in the storage unit 12.

[0053] The crane control unit 115, which serves as the gripper control unit, outputs instruction signals to drive the crane 252 on three axes to grip waste 28 at a desired position with the bucket 253, and instruction signals to scatter the gripped waste 28, and transmits these to the control unit 21. The control unit 21 controls the crane 252 and bucket 253 of the gripper 25 based on the received instruction signals. Note that the crane 252 may also be controlled directly by the crane control unit 115.

[0054] The crane control unit 115 is configured to be able to work in conjunction with the imaging unit 23. The imaging unit 23 can start imaging at the timing when the bucket 253 is moved into the area of the input hopper 331 by the crane control unit 115, and can end imaging at the timing when the bucket 253 is moved out of the area of the input hopper 331.

[0055] The learning unit 116 as a learning means can derive the image analysis model 123 by machine learning using the captured image information of the waste 28 being put into the input hopper 331 as learning input parameters and the image information classified into set types for the acquired image information as learning output parameters. The generated image analysis model 123 is supplied to the learning unit 116 from the computer that performed the machine learning and stored in the memory unit 12.

[0056] The storage unit 12 stores waste information 121, image information 122, and an image analysis model 123. The waste information 121 and the image information 122 are both stored in the storage unit 12 as a searchable database.

[0057] The waste information 121 is information relating to waste 28 stored in the storage pit 26, held by the holding unit 25, or thrown into the input hopper 331. The waste information 121 includes information relating to the classification (also called type) of the waste to be incinerated as waste 28, such as plastic waste, plant waste, and sludge waste.

[0058] The image information 122 includes image information of the waste 28 captured by the imaging camera 231 (232) of the imaging unit 23. The image information 122 includes, for example, video information captured of the waste 28 gripped by the bucket 253 of the gripping unit 25 in a scattered state when the waste 28 is dumped into the dump hopper 331. Note that in this specification, a video is made up of multiple frames of still images, and therefore a video (video) is included in the concept of an image, and video information is included in image information.

[0059] The image analysis model 123 is configured as a learning model in which the image information acquired from the imaging unit 23 that captured the waste 28 in the storage pit 26 is used as an input parameter, and the type (classification) of the waste 28 is used as an output parameter. The learning model is also referred to as a trained model or simply as a model. The image analysis model 123 is generated by machine learning such as supervised learning in which the image information acquired from the imaging unit 23, which captured the waste 28 in a scattered state after being released into the input hopper 331, is used as an input parameter for learning, and the type of waste 28 set by an operator based on the image information is used as an output parameter for learning. Note that various types of machine learning can be used for machine learning, such as deep learning using a neural network.

[0060] (Image analysis model generation method) Here, a method for generating the image analysis model 123 will be described. Fig. 6 is a flowchart for explaining the method for generating an image analysis model executed by the information processing device 10 according to this embodiment. Figs. 7, 8, and 9 are diagrams respectively showing the state in which waste 28 is held in the bucket 253 and about to be thrown into the throwing hopper 331, the state in which it is thrown into the throwing hopper 331, and the state in which it has been separated by type at the waste incineration plant 1 according to this embodiment. In the following description, information is transmitted and received between the respective components via the communication units 13, 22 and the network 2, but a detailed description of this point will be omitted.

[0061] As shown in Fig. 6, in step ST1, the input / output unit 14 of the information processing device 10 acquires, from the imaging camera 231 of the imaging unit 23, imaging information of the bucket 253 and the waste 28 when the waste 28 is being dumped into the dump hopper 331. In this case, as shown in Figs. 7 and 8, the imaging information includes a continuous video of the waste 28 being dumped into the dump hopper 331, from the state immediately before the waste 28 is grasped by the bucket 253 and dumped into the dump hopper 331. In other words, the imaging information includes, as video information, the state in which the waste 28 is released from the bucket 253 and scattered into pieces. The acquired imaging information is input from the input / output unit 14 to the image processing unit 111.

[0062] Next, the process proceeds to step ST2 shown in Fig. 6, where the image processing unit 111 of the control unit 11 extracts at least one still image by cutting out a still image for each frame of the video information in the acquired imaging information. That is, the image processing unit 111 extracts from the video information a still image (see Fig. 7) of the bucket 253 gripping the waste 28, or a still image (see Fig. 8) of the bucket 253 opening and dumping the waste 28 into the feeding hopper 331. These extracted still image data are stored in the memory unit 12 as image information 122.

[0063] Next, the process proceeds to step ST3, where the annotation unit 114 of the control unit 11 performs annotation processing on the still image obtained based on the acquired imaging information. That is, the annotation unit 114 reads out the still image of the image information 122 from the storage unit 12 and outputs the still image from the input / output unit 14. The input / output unit 14 displays the extracted still image. The worker uses the input / output unit 14 to classify and label the extracted still image by type of waste 28. Specifically, as shown in FIG. 9, the worker uses the input / output unit 14 to classify waste 28 classified into a first type, a second type, and a third type in the still image corresponding to FIG. 8, respectively, by a predetermined classification, for example, color coding. In the example shown in FIG. 9, the first type (white: for example, green) is, for example, plastic waste, the second type (dot hatching: for example, red) is, for example, vegetation waste, and the third type (diagonal hatching: for example, brown) is, for example, sludge waste, but this is not necessarily limited thereto. Furthermore, it is possible to perform a predetermined classification (for example, yellow) on the bucket 253. The label set for the still image is stored in the waste information 121 of the storage unit 12 as a processed image associated with the information on the still image.

[0064] Next, the process proceeds to step ST4 shown in FIG. 6, where the learning unit 116 of the control unit 11 reads a still image from the image information 122 of the storage unit 12 and acquires it as a learning input parameter. The learning unit 116 also reads a processed image corresponding to the read still image from the waste information 121 of the storage unit 12 and acquires it as a learning output parameter. That is, the learning unit 116 performs machine learning using the captured image information of the waste 28 being put into the input hopper 331 as a learning input parameter and image information categorized by set type for the acquired image information as a learning output parameter. Here, the machine learning by the learning unit 116 is preferably performed using deep learning, such as a semantic segmentation model, with the learning input parameters and the learning output parameters as input and output data sets. For example, U-Net can be used as the semantic segmentation model.

[0065] Thereafter, the process proceeds to step ST5, where the learning unit 116 stores the learning model generated by machine learning in a readable manner in the storage unit 12 as the image analysis model 123. The image analysis model 123 is read by the type determination unit 112 and used to determine the type of waste 28. The generated image analysis model 123 is supplied to the learning unit 116 from the computer that performed the machine learning, and is stored in the storage unit 12. In this way, the generation process of the image analysis model 123 is executed.

[0066] (Waste Characterization Method) Next, a method for determining the properties of waste according to this embodiment will be described. Fig. 10 is a flowchart for explaining a method for determining the properties of waste 28 executed by the information processing device 10 according to this embodiment. Note that step ST11 is processing by the imaging unit 23 of the waste storage facility 20, and steps ST12 to ST15 are processing by the information processing device 10. Note that in the following description, information is transmitted and received between the respective components by inputting and outputting signals, or by transmitting and receiving via the communication units 13, 22 and the network 2, but a detailed description of this point will be omitted.

[0067] 10 , first, in step ST11, the imaging unit 23 of the waste storage facility 20 images the waste 28 being thrown into the input hopper 331 and outputs the image as video information. The video information captured by the imaging unit 23 is transmitted to the information processing device 10. The control unit 11 of the information processing device 10 stores the acquired video information in the memory unit 12 as image information 122. Note that in step ST11, the control unit 11 of the information processing device 10 may control the imaging unit 23 via the control unit 21 of the waste storage facility 20 to capture an image of the waste 28 being thrown into the input hopper 331.

[0068] Next, the process proceeds to step ST12, and in the information processing device 10, the image processing unit 111 of the control unit 11 performs image processing on the acquired video information to cut out and extract still images. As a result, a still image of the waste 28 being held in the bucket 253 (see, for example, FIG. 7) and a still image of the waste 28 being released from the bucket 253 (see, for example, FIG. 8) are cut out. The extracted still images are stored in the storage unit 12 as image information 122. Furthermore, when acquiring these still images, it is possible to not only extract the still images but also perform image processing such as brightness correction, chromaticity correction, and partial image deletion on the still images.

[0069] Next, the process proceeds to step ST13, where the type determination unit 112 of the control unit 11 reads out the image analysis model 123 from the storage unit 12. The type determination unit 112 of the control unit 11 also reads out and acquires a still image of the bucket 253 and waste 28 captured by the imaging unit 23 from the image information 122 of the storage unit 12. The type determination unit 112 inputs the acquired still image as an input parameter to the image analysis model 123. The image analysis model 123 recognizes a group of pixels forming a characteristic category for each pixel of the input still image, and outputs, as an output parameter, determination result information that determines the waste type for each pixel of the still image. This allows the type determination unit 112, which performs processing in accordance with the image analysis model 123, to divide the still image of the waste 28 (see, for example, FIG. 8) into multiple regions at the pixel level for each type of waste 28.

[0070] Then, the process proceeds to step ST14, where the waste information generation unit 113 generates numerical information for each type of waste 28 captured in the still image, divided into multiple regions at the pixel level for each type, based on the determination result information for the waste type obtained from the type determination unit 112.

[0071] Here, numerical information can be derived, for example, by using the area, brightness distribution, and pigment distribution of each waste type output as a result of image analysis, the particle size of the waste obtained based on the area of each waste region, and the shape of each waste region obtained by fractal analysis. When capturing images of the waste 28 using the imaging cameras 231 and 232, still images are generated by cutting out the captured video at regular intervals. By performing image analysis on each still image obtained by cutting out the captured video, it is possible to obtain information on changes in the waste region over time. This makes it possible to obtain information such as the movement speed and viscosity of the falling waste 28. Waste information can be generated by analyzing at least one piece of numerical information obtained from image analysis, or by combining multiple pieces of numerical information. Statistical methods such as regression analysis and machine learning methods such as decision tree analysis can be used as analytical methods.

[0072] Next, in step ST15, the type determination unit 112 sets a label corresponding to the type based on the waste type determination result information generated for each area divided at the pixel level in the still image, and generates and outputs a processed image (e.g., see FIG. 9) in which the classified image analysis result is superimposed on the target still image (e.g., see FIG. 8). Also, in step ST14, the waste information generated by the waste information generation unit 113 is output. This allows the control unit 11 to visually indicate the image type determination result for the waste 28 captured in the still image, i.e., the waste 28 in the state it has been placed in the input hopper 331, and makes it possible to derive waste information such as the content rate for each type.

[0073] FIG. 11 shows the measured values of the time change in the plastic waste ratio among the types of waste 28 determined by the above-described property determination process and the lower heating value (Hu value) after, for example, 60 minutes, until the waste 28 input into the input hopper 331 is burned in the incinerator 33. From FIG. 11, it can be seen that there is a close correlation between the image analysis results of the property determination process and the plastic waste ratio. In other words, it is possible to derive the proportion of plastic waste contained in the waste 28 input into the input hopper 331, and it can be seen that indicators such as the plastic waste ratio are correlated with the lower heating value of the input waste 28. Therefore, it can be seen that the property determination process for waste 28 according to the above-described embodiment makes it possible to accurately determine the type (property) of the waste 28 when it is input into the input hopper 331.

[0074] (First Modification) Next, a first modified example of the above-described embodiment will be described. Fig. 12 is a top view illustrating the installation positions of a plurality of, for example, two, imaging cameras installed in a feeding hopper 331 according to a first modified example of the present embodiment.

[0075] 12, two imaging cameras 231, 232 are provided above or above the input hopper 331. In this case, it is preferable to arrange them so that the imaging camera 232 can capture images of areas that would otherwise be blind spots for the imaging camera 231, and so that the imaging camera 232 can capture images of areas that would otherwise be blind spots for the imaging camera 232. In a modified example, the two imaging cameras 231, 232 are each arranged so that the angle at which they capture images of the waste 28 is an angle φ along a horizontal plane, with the position at which the crane 252 releases the waste 28 as the center. Here, the angle φ is typically greater than 0° and less than or equal to 180° (0°<φ≦180°), preferably greater than or equal to 45° (45°≦φ), and more preferably greater than or equal to 90° (90°≦φ). Furthermore, when installing multiple imaging cameras such as imaging cameras 231 and 232, it is preferable to install them at positions and angles φ that are unlikely to create blind spots between imaging cameras 231 and 232, in order to prevent blind spots between the imaging cameras.

[0076] (Second Modification) Next, a second modified example of the embodiment described above will be described. Fig. 13 is a block diagram showing the information processing device according to the second modified example in Fig. 1. Fig. 13 corresponds to Fig. 5 in the embodiment. Figs. 14 and 15 are flowcharts for explaining the learning method and analysis method in the waste property determination method according to the second modified example. The flowcharts shown in Figs. 14 and 15 in the second modified example correspond to the flowcharts shown in Figs. 6 and 10 in the embodiment, respectively. Furthermore, the learning method and analysis method according to the second modified example described below will be described as a method for using an image classifier based on CNN, such as VGG or ResNet, as the image analysis model.

[0077] As shown in FIG. 13, the information processing device 10 according to the second modification has a property determination unit 117 instead of the type determination unit 112, and a labeling unit 118 instead of the annotation unit 114. In the second modification, the functions of the image processing unit 111, the waste information generation unit 113, the crane control unit 115, the learning unit 116, the property determination unit 117, and the labeling unit 118 are performed by executing various programs loaded by the control unit 11. The various programs also include programs that realize artificial intelligence and trained models capable of realizing the processing according to this embodiment. The functions of the property determination unit 117 and the labeling unit 118 will be described in detail below.

[0078] (Image analysis model generation method) As shown in FIG. 14, steps ST21 and ST22 are similar to steps ST1 and ST2 shown in FIG. 6 according to the above-described embodiment. That is, in step ST21, the input / output unit 14 of the information processing device 10 acquires imaging information, and in step ST22, the image processing unit 111 of the control unit 11 extracts at least one still image from the acquired imaging information. At this time, it is possible to simply extract a still image from a moving image, or further, it is possible to perform image processing such as brightness correction and deletion of unnecessary portions on the extracted still image. The extracted still image data, which has been subjected to image processing as necessary, is stored in the storage unit 12 as image information 122.

[0079] Next, the process proceeds to step ST23, where the labeling unit 118 of the control unit 11 performs a labeling process on the still image data obtained based on the acquired imaging information. That is, the labeling unit 118 assigns a property value to be predicted to the still image. Here, the property value is, for example, an actual value obtained by monitoring the state in which the waste 28 captured in the acquired still image is fed into the incinerator 33 through the feeding hopper 331 after a predetermined time, burned, and discharged as ash or exhaust gas. Specifically, for example, the property value is the lower heating value (Hu) value after the time required for the waste 28 to be burned after being fed into the feeding hopper 331 has elapsed.

[0080] Next, the process proceeds to step ST24, where the learning unit 116 performs learning using an image analysis model based on the acquired still image data and the attribute values labeled on the still image data. That is, the still image is used as the learning input parameter, and the labeled attribute values are used as the learning output parameter to learn an image analysis model such as CNN. Here, not only still image data but also measurement values contained in sensor information can be added as learning input parameters. The learning unit 116 performs machine learning using the learning input parameters and learning output parameters as input / output data sets. Thereafter, the process proceeds to step ST25, where the learning unit 116 stores the learning model generated by machine learning as the image analysis model 123 in the storage unit 12 so that it can be read by the attribute determination unit 117. In this manner, the generation process of the image analysis model 123 is executed.

[0081] (Waste Characterization Method) Next, a description will be given of a method for determining the properties of waste according to a second modified example of this embodiment. Determining the properties of waste according to the second modified example is carried out as shown in FIG.

[0082] That is, in step ST31, the imaging unit 23 images the waste 28 and transmits video information to the information processing device 10. Here, the video information is an image of the waste 28 captured by the imaging unit 23 in a state where the properties of the waste 28 are unknown. The video information is stored in the memory unit 12 as image information 122. Next, the process proceeds to step ST32, where the image processing unit 111 cuts out and extracts a still image from the video information. The still image of the waste 28 in which the properties of the waste 28 are unknown is stored in the memory unit 12 as image information 122.

[0083] Next, the process proceeds to step ST33, where the property determination unit 117 reads out a still image from the image analysis model 123 and the image information 122. The property determination unit 117 inputs the acquired unknown still image as an input parameter into the image analysis model 123. Next, the process proceeds to step ST34, where the property determination unit 117 outputs a predicted value of the property of the waste 28 used in the labeling process during learning. As a result, the control unit 11 determines the property of the waste 28.

[0084] According to the embodiment described above, the image analysis model 123 is used to determine the properties of the waste 28 put into the input hopper 331 and classify the type of waste 28, thereby making it possible to grasp the properties of the waste 28 before it is input into the incinerator 33. As a result, the transport time (e.g., 60 minutes) from the time the waste 28 is input into the input hopper 331 until the input waste 28 is incinerated in the incinerator 33 can be calculated in advance, and the combustion state inside the incinerator 33 after the transport time from the time the waste 28 is input into the input hopper 331 can be calculated using the combustion control device 31, etc., so that fluctuations in the situation inside the incinerator 33 can be predicted, enabling the waste incineration plant 1 to be operated stably.

[0085] Although one embodiment of the present invention has been specifically described above, the present invention is not limited to the above-described embodiment, and various modifications based on the technical concept of the present invention are possible. For example, the numerical values given in the above-described embodiment are merely examples, and different numerical values may be used as necessary. The present invention is not limited by the descriptions and drawings that form part of the disclosure of the present invention according to this embodiment.

[0086] For example, in the embodiment described above, at least one of the imaging cameras 231, 232 constituting the imaging unit 23, preferably at least one of the multiple imaging cameras 231, 232, may be provided, for example, at the bottom of the crane girder (not shown). Furthermore, it is preferable that the at least one imaging camera 231, 232 is configured to be able to capture images of as much of the waste material 28 as possible among the waste material 28 held in the bucket 253. In this case, if necessary, at least one of the multiple imaging cameras 231, 232 may be disposed on the ceiling or elsewhere other than approximately the center of the input hopper 331.

[0087] For example, in the embodiment described above, the waste information generation unit 113 classifies the waste 28 into first to third types, but this is not necessarily limited to three types, and the waste may be classified into two types, or into four or more types, and the number of types is not limited.

[0088] Furthermore, when artificial intelligence is used in the above-described embodiment, deep learning using a neural network is used as an example of machine learning, but machine learning based on other methods may also be performed. For example, other supervised learning methods such as support vector machines, decision trees, naive Bayes, and k-nearest neighbor methods may also be used. Furthermore, semi-supervised learning may also be used instead of supervised learning. Furthermore, the image analysis method is not limited to one type, and it is also possible to combine the output results obtained by image analysis using semantic segmentation employed in the embodiment with the output results obtained by image analysis using CNN employed in the second modified example.

[0089] In addition, in one embodiment, the above-mentioned "unit" can be read as "circuit" etc. For example, a control unit can be read as a control circuit.

[0090] In the explanation of the flowcharts in this specification, the order of processing between steps is clearly indicated using expressions such as "first," "then," and "continue," but the order of processing required to implement this embodiment is not uniquely determined by these expressions. In other words, the order of processing in the flowcharts described in this specification can be changed within a consistent range.

[0091] Further advantages and modifications will readily occur to those skilled in the art. The disclosure in its broader aspects is not limited to the specific details and representative embodiments shown and described above. Accordingly, various modifications may be made without departing from the spirit or scope of the general inventive concept as defined by the appended claims and their equivalents. [Explanation of symbols]

[0092] 1. Waste incineration plant 2 Network 3. Waste treatment facilities 10. Information processing equipment 11,21 Control unit 12 Storage section 13,22 Communications Department 14 Input / output section 20 Waste storage facility 23 Imaging unit 24,32 Sensor section 25 Gripping part 26 Storage pit 27 Platform 28 Waste 30 Waste incineration facility 31 Combustion control device 33 Incinerator 111 Image processing unit 112 Type determination section 113 Waste Information Generation Department 114 Annotation Section 115 Crane control unit 116 Learning Department 117 Property Judgment Department 118 Labeling Department 121 Waste Information 122 Image Information 123 Image Analysis Model 231,232 Imaging camera 252 Crane 253 Bucket 331 Feeding hopper

Claims

1. a control unit having hardware, The control unit image information is acquired from an imaging unit having at least one imaging camera capable of imaging the waste when the waste is being put into an input unit configured to be able to input the waste and supply the waste to an incinerator; extracting a still image of the waste included in the image information; inputting the still image as an input parameter into an image analysis model; classifying a plurality of types of waste contained in the waste in the still image by the types and outputting the classified waste as output parameters of the image analysis model; The image analysis model is a learning model generated by machine learning using still images extracted from video information capturing the state of waste being put into the input section as learning input parameters, and images of the waste captured in the still images classified by type as learning output parameters. Information processing device.

2. The imaging unit is provided at a height that is in the range of -45 degrees to 45 degrees from the horizontal direction with respect to the height at which the gripping unit capable of gripping and releasing the waste throws the waste into the throwing unit. The information processing device according to claim 1 .

3. the imaging unit includes a plurality of imaging cameras, At least two of the plurality of imaging cameras are arranged so that the angle at which the waste is imaged is within a range of more than 0° and not more than 180° along a horizontal plane, with the position at which the gripping unit capable of gripping and releasing the waste releases the waste as the center. The information processing device according to claim 1 .

4. the imaging unit includes a plurality of imaging cameras, At least two of the plurality of imaging cameras are arranged so that the angle at which the waste is imaged is within a range of 90° to 180° along a horizontal plane, with the center being the position at which the gripping unit capable of gripping and releasing the waste releases the waste. The information processing device according to claim 1 .

5. The image analysis model is a learning model generated by a semantic segmentation algorithm. The information processing device according to claim 1 .

6. classifying a plurality of types of waste contained in the waste in the still image by the types as output parameters of the image analysis model; generating an image analysis result by color-coding the classified types; A processed image is generated by superimposing the image analysis result on the still image, and the processed image is output. The information processing device according to claim 1 .

7. An information processing method executed by a control unit having hardware, The waste is then fed into an incinerator through an incinerator. The incinerator is then fed into an incinerator through an incinerator. extracting a still image of the waste included in the image information read from the storage unit; inputting the still image as an input parameter into an image analysis model; classifying a plurality of types of waste contained in the waste in the still image by the types and outputting the classified waste as output parameters of the image analysis model; The image analysis model is a learning model generated by machine learning using still images extracted from video information capturing the state of waste being put into the input section as learning input parameters, and images of the waste captured in the still images classified by type as learning output parameters. Information processing methods.

8. classifying a plurality of types of waste contained in the waste in the still image by the types as output parameters of the image analysis model; generating an image analysis result by color-coding the classified types; A processed image is generated by superimposing the image analysis result on the still image, and the processed image is output. The information processing method according to claim 7.

9. A program to be executed by a control unit having hardware, The control unit The waste is then fed into an incinerator through an incinerator. The incinerator is then fed into an incinerator through an incinerator. extracting a still image of the waste included in the image information read from the storage unit; inputting the still image as an input parameter into an image analysis model; classifying a plurality of types of waste contained in the waste in the still image by the types and outputting the classified waste as output parameters of the image analysis model; The image analysis model is a learning model that has been generated in advance by machine learning, using as learning input parameters still images extracted from video information capturing the state of waste being put into the input section, and as learning output parameters images of the waste captured in the still images classified by type. program.

Images