Device, system, program and method for estimating quality of waste

Abstract

To provide a device for estimating quality of waste in a refuse pit by using a learning model in the refuse pit in which waste having different quality is mixed.SOLUTION: A device includes: a teacher data generation section generating teacher data associated with a captured image of waste stored in a refuse pit; a model construction section constructing a model through learning using the teacher data; and an estimation section inputting data on a new captured image of the waste stored in the refuse pit to the model and acquiring a value indicating quality of the waste corresponding to the new image.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an apparatus, system, program, and method for estimating waste quality in a waste treatment plant. The present invention relates to a ram and a method. [Background technology]

[0002] Waste treatment plants process a variety of waste materials, including household waste, crushed bulky waste, pruning branches, and sludge. Waste with various qualities is dumped into the garbage pit. However, the waste that is put into the incinerator is If the quality of the waste being incinerated changes suddenly, the temperature inside the incinerator may change suddenly during incineration, or However, harmful gases and substances such as dioxins can be generated, which can cause environmental problems.

[0003] Therefore, for example, in Patent Document 1, general waste and different waste thrown into a garbage pit are distinguished by color. The system can identify the waste more accurately and control the crane to mix the waste in the pit, thereby achieving uniform quality of the waste. are.

[0004] However, in the technology described in Patent Document 1, for example, kitchen waste and plus It is also possible to distinguish between garbage that is different in quality but similar in color, such as plastic bags, and between garbage that is the same in quality but similar in color, such as bedding. Identifying the quality of the variable waste was difficult. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 5025120 Specification Summary of the Invention [Problem to be solved by the invention]

[0006] This technology was developed in consideration of the above points, and is suitable for waste materials with different qualities but similar colors, or waste materials with similar qualities. Even if waste of the same color is mixed, the quality of the waste can be estimated. One of the objects is to provide an apparatus, a system, a program, and a method. [Means for solving the problem]

[0007] [Form 1] According to form 1, the waste stored in the garbage pit is associated with an image captured. a training data generating unit for generating training data based on the training data; The model construction section constructs a model, and the new image captures the waste stored in the garbage pit. The data of the new image is input to the model to represent the waste quality corresponding to the new image. and an estimator for obtaining the value.

[0008] [Mode 2] According to mode 2, in the device according to mode 1, the estimation unit further A new image of the waste material is divided into a plurality of blocks, and the new image is A value representing the quality of the corresponding waste is output, and the output value representing the quality of the waste is An inference map is generated corresponding to each of the above.

[0009] [Feature 3] According to feature 3, in the device according to feature 2, the device further comprises: Based on the logic map, instructions are sent to the crane control device that controls the crane, or An instruction unit is provided that generates at least one of instructions to a combustion control device that controls the incinerator. can.

[0010] [Feature 4] According to feature 4, in the device according to feature 3, the crane control device The instruction is to move the waste in the garbage pit to the crane, and the instruction is to move the waste in the garbage pit to the crane. The instructions to the control device are those necessary to burn the waste that has been put into the incinerator. .

[0011] [Feature 5] According to feature 5, in the device of any one of features 1 to 4, The data are values ​​that indicate the characteristics of the waste identified based on the operating history of the waste treatment plant. and labels that workers have used to classify the quality of waste based on image data of the waste in the garbage pit. It is collected from at least one of them.

[0012] [Mode 6] According to Mode 6, in the device of any one of Modes 1 to 5, The value representing the quality of waste is an index showing the combustibility of said waste.

[0013] [Embodiment 7] According to embodiment 7, a waste treatment plant system is provided, in which waste is stored in a garbage pit. a teacher data generation unit that generates teacher data associated with an image of the waste to be captured; a model construction unit that constructs a model by learning using the training data; The new image of the waste stored in the tank is divided into multiple blocks, and each block is The data of the new image is input to the model to determine the waste quality corresponding to the new image. an estimation unit that generates an inference map in which a value representing Based on the plan, instructions are given to the crane control device that controls the crane, or the incinerator and an instruction unit that generates at least one of instructions to the combustion control device that performs the control. A system is provided.

[0014] [Form 8] According to Form 8, the quality of waste stored in the garbage pit of the waste treatment plant A method for estimating the amount of waste stored in a garbage pit, which is associated with an image of the waste stored in the garbage pit. generating training data based on the training data; and The steps to construct the system and new image data of waste stored in the waste pit , input into the model to obtain a value representing the quality of the waste corresponding to the new image. and a method is provided.

[0015] [Mode 9] According to mode 9, the method according to mode 8 is provided in the waste treatment plant. A program for causing the processor to execute the program is provided.

[0016] [Form 10] According to form 10, the device for controlling the operation of a waste treatment plant is Estimating the quality of waste corresponding to images of waste stored in a waste pit in a processing plant a data structure for use in operating the waste treatment plant, the data structure comprising: The value representing the quality of waste generated from history and the waste corresponding to the value representing the quality of the waste and the device includes training data including an image, and the device learns the previous image by using the training data. The quality of the waste corresponding to the new image of the waste stored in the waste pit is estimated. A data structure is provided. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a schematic diagram of a waste treatment plant according to one embodiment; [Figure 2] 1 is a schematic configuration diagram of a waste treatment plant system according to an embodiment. [Figure 3] 1 is a functional configuration diagram of an information processing device of a waste treatment plant according to an embodiment; [Figure 4]This is an example of an inference map showing the correspondence between output data and positions within the garbage pit. [Figure 5] 1 is a flowchart illustrating the operation of a waste treatment plant system according to one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0019] FIG. 1 shows a schematic diagram of a waste treatment plant according to one embodiment. In FIG. 1: 1 is an incinerator that burns waste, 2 is a waste heat boiler, 3 is a pit that stores garbage, and 4 is a hopper, 5 is a crane for transferring waste from pit 3 to hopper 4, and 6 is a crane for storing waste in pit 3. In FIG. 1, 21 is a platform. The garbage collected by the garbage truck 22 from the platform 21 is put into the garbage pit 3. It will be deployed.

[0020] FIG. 2 shows a block diagram of a system 100 for controlling the operation of the waste treatment plant shown in FIG. The waste treatment plant system 100 uses the image data captured by the imaging device 6 to A value representing the quality of waste is estimated, and a waste treatment plan is established based on the estimated value representing the quality of waste. The waste treatment plant system 1 is configured to control the operation of the plant. 00 includes an information processing device 200, an imaging device 6 for taking pictures of the inside of a garbage pit 3 (shown in FIG. 1), The information processing device 200 includes a crane control device 110 and a combustion control device 120. For example, the imaging equipment can be connected to a network such as an on-site LAN installed in a waste treatment plant. The device 6, the crane control device 110, and the combustion control device 120 are connected to each other so as to be able to communicate with each other. The information processing device 200 is, for example, a personal computer or a workstation. It may be configured as an application, a server device, or a portable computer such as a tablet terminal. This configuration is just an example, and the configuration to which the present invention can be applied is shown in FIG. For example, the imaging device 6, the crane control device 110, and the combustion control device There may be two or more information processing devices 120 and 200.

[0021] The image capturing device 6 captures the surface of the waste piled up in the garbage pit 3 and outputs an image data of the inside of the pit 3. The image capturing device 6 is a device for capturing images of the shape and color of waste, for example. an RGB camera that captures near-infrared images of the waste; an infrared camera that captures near-infrared images of the waste; It is a 3D camera or RGB-D camera that captures the above.

[0022] The information processing device 200 includes a processor 202, a memory 204, and , a communication interface 206, and storage 208. The information processing device 200 detects the location of the garbage pit 3 based on the image data of the garbage pit 3 transmitted from the imaging device 6. The information processing device 200 generates training data to be given to the learning model. The details of this function will be described later with reference to FIG.

[0023] The processor 202 reads a program stored in the memory 204 and executes the program accordingly. The processor 202 is configured to execute the process stored in the memory 204. By executing the program, the functions of the processes described below are realized. The processor 202 includes a CPU (Central Processing Unit), MPU (Micro Processor Unit), FPGA (Field-Pro It is realized as a device such as a programmable gate array.

[0024] The memory 204 temporarily stores programs and data. The data is loaded from the storage 208. and data generated by the processor 202 and loaded from the storage 208. In one embodiment, the memory 204 includes RAM (Random Access Memory). The data stored in the memory 204 is realized as a volatile memory such as a volatile storage memory (SS Memory). The data includes image data of the waste photographed by the imaging device 6, and a This includes teacher data generated by the

[0025] The communication interface 206 is connected to the image capture device 6, the crane control device 110, and the combustion control device 110. Signals are communicated between the device 120 and the information processing device 200. The communication interface 206 receives the image data output from the image capture device 6. In this embodiment, the communication interface 206 communicates instructions generated by the processor 202 to The signal is sent to the crane control device 110 or the combustion control device 120.

[0026] The storage 208 permanently stores programs and data. For example, ROM (Read-Only Memory), hard disk drive, flash The program stored in the storage 208 is realized as a non-volatile storage device such as a flash memory. The program generates training data based on image data of waste photographed by the imaging device 6, for example. a program for generating the data, a crane control device 110, or a combustion control device 12 The storage 208 includes a program for providing instructions to the waste disposal process. The operation history of the waste treatment plant is stored in chronological order. A plurality of sensors and other devices are attached to a treatment plant, for example, an incinerator 1 (shown in FIG. 1). The temperature sensor detects the temperature inside the storage device. The page 208 includes a database 209, which contains, for example, The process data obtained based on the operation history of the processing plant is stored in chronological order. The process data may be stored in memory 204.

[0027] The crane control device 110 is a device that controls the operation of the crane 5 (shown in FIG. 1). The crane control device 110 controls the crane in accordance with the instructions sent from the information processing device 200. 5 to perform the operation of stirring the waste in the garbage pit 3, or to The waste in the garbage pit 3 is conveyed to the hopper 4 (shown in Figure 1). A portion of the waste piled up in a block in the waste pit is picked up by crane 5 and transported to the waste pit. The waste is then moved to another block in Block 3, or the waste picked up by Crane 5 is then returned to the same block. By repeatedly stirring the waste, the waste in the garbage pit 3 is uniformly mixed. This allows the waste to be burned uniformly in the incinerator 1. can.

[0028] The combustion control device 120 is a device that controls combustion in the incinerator 1 (shown in FIG. 1). The control device 120 controls the incinerator 1 in accordance with instructions sent from the information processing device 200. The burning time and combustion temperature are controlled, or the amount of air sent into the incinerator 1 is controlled.

[0029] FIG. 3 is a block diagram showing the functional configuration of an information processing device 200 according to an embodiment of the present invention. The information processing device 200 according to this embodiment includes a training data generating unit 220 and a model The system includes a rule constructing unit 230, an image acquiring unit 240, an estimating unit 250, and an instructing unit 260. The units 220 to 260 are the computer programs stored in the memory 204 by the processor 202 shown in FIG. This represents a function that is realized by reading and executing a program.

[0030] According to this embodiment, image data of waste included in the operation history of the waste treatment plant, etc. Generate training data from the waste data and collect multiple sets of the image data and training data. By collecting these and applying machine learning to them, a learning model is constructed. Input image data of the waste to be estimated into the learning model and obtain the output (inference results). Based on the output, an inference map is generated. Based on the inference map, a crane control 2, and the combustion control device 120 (shown in FIG. 2). The operation of each of the units 220 to 260 will be described in detail below.

[0031] The teacher data generation unit 220 generates teacher data to be given to the learning model. The generating unit 220 receives the data stored in a predetermined database 209 (shown in FIG. 2) in chronological order. Training data is generated from the past or current operating history of the waste treatment plant. The database 209 stores image data of the waste in the garbage pit 3 captured by the imaging device 6. The waste image data and the process data corresponding to the waste image data are stored in chronological order. The image data generating unit 220 obtains image data of the waste in the garbage pit 3 from the database 209. The process data corresponding to the data is read out, and the process data is used as training data for learning. The image data of the waste in the garbage pit 3 and the process data are generated, for example, as follows: These are associated with each other based on the time they were acquired. In other words, image data of past waste includes the following: Past process data is associated with the training data generated from the process data. It includes values ​​that represent the quality of waste. Values ​​that represent the quality of waste are based on the combustibility or flammability of the waste. The process data is an index that indicates the difficulty of the waste treatment plant. Based on the data showing the characteristics of the collected waste and the operation history of the waste treatment plant, workers and a label classifying the quality of the waste.

[0032] The data indicating the characteristics of the waste may be, for example, the weight of the waste placed in the pit 3 (kg· m / s 2 ), density (kg / m 3 ), the moisture content of the waste placed in pit 3 (k g), or the amount of heat generated when burning waste (kJ / kg). The low density, low moisture content, and high calorific value of the waste mean that it is easily combustible. Generally, the lower the density, the easier it is to burn waste. The volume of waste that crane 5 can grab is almost the same. Therefore, if the weight of the waste that the crane 5 grabs is measured, the combustibility of the waste can be determined. It can be determined.

[0033] For example, the database 209 may contain information based on the past operation history of the waste treatment plant. The data indicating the characteristics of the waste is recorded in advance. The monthly heat output of a waste treatment plant over a specified period (e.g., three years) is included in the The information is stored in advance. By referring to the history information, the waste shown in the image of the waste in the garbage pit 3 can be The actual calorific value of the waste when it is dumped into pit 3 can be determined.

[0034] In this embodiment, the image data of the waste in the garbage pit 3 is shown in the image of the waste. The waste was collected in accordance with the data showing the characteristics of the waste identified when it was dumped into the waste pit 3. Then, the training data generating unit 220 generates data indicating the characteristics of the identified waste. The data is then shaped to generate training data that represents the quality of the waste, corresponding to the waste image data. Complete.

[0035] On the other hand, the labels that classify the quality of waste are stored in the database 209. Based on the image data, workers visually classify the quality of waste shown in the image. For example, if a worker determines that the waste shown in the image is made up of high-quality waste, If it is determined that the waste is composed of standard waste, it is assigned the label "H". If it is determined that the waste is composed of low-quality waste, it is assigned the label "L". The assigned label is input to the information processing device via an input interface (not shown). The data is input to the device 200 and stored in the database 209.

[0036] In this embodiment, image data of the waste in the garbage pit 3 captured by the imaging device 6 Labels that workers visually classify the quality of waste are collected as training data.

[0037] In this way, the teacher data generating unit 220 generates the waste pixilation data from the operation history of the waste treatment plant. and acquiring process data corresponding to the image of the waste in the waste container 3, and The training data generating unit 220 generates data on the image of the waste and the image of the waste. We collect multiple sets of training data that contain values ​​representing the quality of waste corresponding to the image, and create training data. The data is stored in memory 204.

[0038] The training data generating unit 220 uses images of past waste stored in the database 209 The image data is divided into a plurality of blocks, and for each divided block, a corresponding The process data may be acquired, and the teacher data may be generated from the acquired process data. In this case, the teacher data generating unit 220 may, for example, A block number is assigned to each waste, and a value representing the quality of the waste is stored in the memory 204 together with the block number. Therefore, the teacher data generating unit 220 can determine which block in the pit 3 contains the waste. It is possible to generate a map of training data that shows how flammable something is.

[0039] The model construction unit 230 constructs a model (function) by machine learning the generated training data. The model is designed to generate the correct output when a new input is received. The model construction unit 230 has a predetermined function y=f(x, θ), where , input x is image data of waste in the garbage pit 3, output y is a value representing the quality of waste, θ are the internal parameters of this function. The model construction unit 230 calculates the set of input x and output y. By giving multiple values ​​and performing machine learning, the internal parameter θ The input x given for machine learning is generated by the training data generation unit 220. The image data of the waste in the garbage pit3 collected from the operation history of the waste treatment plant was The input x is the training data, and the output y is the training data corresponding to the input x. The image data of the waste in the garbage pit 3 given as The model construction unit 230 uses the data of these various images to By machine learning multiple sets of training data corresponding to the various image data, ,Find the relationship between the image data and the training data, and adjust the internal parameter θ of the model. This allows the model to adapt to new inputs that are different from the inputs x given so far. Even if a new input is input, the correct output is generated. Algorithms include linear regression, Boltzmann machines, neural networks, and support base Machine learning, Bayesian networks, sparse regression, decision trees, and random forests At least one of the following techniques is used: statistical estimation, reinforcement learning, and deep learning.

[0040] The image acquisition unit 240 acquires an image of the waste in the garbage pit 3 for input into the constructed model. The image acquisition unit 240 acquires image data of the waste in the garbage pit 3 from the imaging device 6. Image data is periodically transmitted or a request for the crane 5 to dump waste into the pit 3 is transmitted. As a trigger, it receives and stores in memory 204.

[0041] The estimation unit 250 acquires new waste image data from the memory 204 and uses the constructed model. The image data of the new waste is input to the data processor, and output data is obtained. , a value representing the quality of the waste corresponding to the new image data, e.g., a value representing the characteristics of the waste. This is the data.

[0042] The estimation unit 250 further calculates the acquired output data and the data in the garbage pit 3 of the output data. The estimation unit 250 may generate an inference map showing the correspondence between the position in the image acquisition unit 2 and the position in the image acquisition unit 2. 40, the new image acquired by dividing it into a plurality of blocks, Each new image data is input to the constructed model and output block by block. 4 shows a diagram of a garbage pit 3, which is constructed by dividing the surface of the garbage pit 3 into a plurality of blocks 402. In the example of FIG. 4, the surface of the garbage pit 3 is 4 in width and 1 in length. The inference map 400 is divided into two parts, and consists of a total of 48 blocks 402. The estimation unit 250 generates a three-dimensional inference map by using a map generated in the past. It can also be achieved.

[0043] Each block 402 constituting the inference map 400 shows output data as an example. In the example shown in Figure 4, the output data is an index showing the combustibility of waste (unitless in Figure 4). For example, the larger the value of the index shown in Figure 4, the easier it is for the waste to burn. The output data written to the block 402 is not limited to that shown in FIG. 4 and may be other data, for example: Weight of waste (kg m / s 2 ), density (kg / m 3 ), moisture content of waste (kg), heat It may be the amount (kJ / kg) or any combination of these. There may be multiple pieces of output data shown in each block 402 that make up the step 400. The map 400 is updated every time there is a change in the image data acquired by the image acquisition unit 240. The inference map 400 is updated / recorded at regular intervals. may be visually displayed.

[0044] Furthermore, each block 402 constituting the inference map 400 contains not only the output data itself but also Alternatively, values ​​(labels, flags, etc.) extracted based on the output data may be displayed. For example, each block 402 is shown with a label classified based on the size of the output data. For example, if the output data is the moisture content of waste, the one with the highest moisture content will be labeled. "L (or low-quality waste)" and those with average moisture content are labeled "M (or standard waste)." " and those with a low moisture content are assigned the label "H (or high-quality waste)". For example, each block 402 indicates a flag determined based on the size of the output data. For example, if the output data is the calorific value, the calorific value of the waste is above a certain level and is incinerated. If it is judged to be suitable for incineration, the flag is set to "OK" and the calorific value is below a certain level. If it is judged that it is not suitable for feeding into furnace 1, it is assigned a flag of "NG". If there are multiple pieces of output data corresponding to the block 402, then from these multiple pieces of output data, The values ​​shown in each block 402 may be newly extracted.

[0045] With this technology, when new image data is input, the correct value (waste) corresponding to it is generated. The model was built to output a value representing the quality of the waste material, and new image data was then used to Therefore, new image data of the waste is collected and these images are used to estimate the quality of the waste. By inputting image data into the learning model, the quality of waste corresponding to new image data can be estimated. Furthermore, it is possible to separate waste materials that are similar in color but have different qualities, or waste materials that do not have a consistent color. Even if waste of the same quality is mixed in Pit 3, the quality of the waste can be estimated from color alone. This technology can estimate the quality of waste with higher accuracy than the conventional method. According to the study, the quality of waste is estimated mechanically using a learning model, so there is no need for skilled workers. It is possible to estimate the quality of waste without visual inspection, or to estimate the quality of waste without visual inspection. This can support the accuracy of the judgments made by workers. By periodically performing additional learning and re-learning using data and corresponding process data, It is also possible to respond to changes in the quality of waste over time.

[0046] In addition, according to this technology, the inside of the garbage pit 3 is divided into a plurality of blocks 402, and block 4 For each waste type, an inference map 400 is generated, showing values ​​representing the quality of the waste. Alternatively, the information processing device 200 may refer to the inference map 400 to determine the location of the garbage pit 3. It is possible to grasp the quality of the waste for each block 402. This allows us to understand the spatial distribution of waste quality.

[0047] The instruction unit 260 controls the crane control device 110 (shown in FIG. 2) based on the inference map 400. More specifically, the instruction unit 260 gives instructions to the user based on the inference map 400. instructions indicating from which block to which block in the waste pit 3 the waste should be moved; Or which block of waste in the garbage pit 3 is to be dumped into the incinerator 1 (shown in Figure 1) The crane control device 110 generates an instruction indicating whether the crane should be moved or not, and sends it to the crane control device 110. The operator 0 controls the crane 5 to move the waste in the garbage pit 3 according to the instructions. The display unit 260 uses the values ​​shown in the inference map 400 to determine whether the quality of the waste in the pit 3 is uniform. or so that the quality of the waste is close to that of the waste previously fed into incinerator 1. Instructs the movement of.

[0048] The instruction unit 260 also controls the combustion control device 120 (shown in FIG. 2) based on the inference map 400. More specifically, the instruction unit 260 instructs which blocks in the inference map 400 The input information of whether the waste of the block was input to the incinerator 1 and the value shown in the block, for example, For example, the incinerator 1 generates instructions necessary for burning the waste that has been input using the output data. The combustion control device 120 then controls the incinerator in accordance with the instructions. The incinerator 1 is designed to have a combustion temperature, combustion time, and air volume that are appropriate for the quality of the waste fed into it. Control combustion.

[0049] The present invention is not limited to the above-described embodiment. 260 is not the information processing device 200 but another device, for example, the crane control device 110 In this case, the inference map 400 generated by the information processing device 200 may be The crane control device 110 then selects the location on the inference map 400. Generates instructions to input waste in the block into incinerator 1 and controls the operation of crane 5. The crane control device 110 controls the amount of waste thrown into the waste incinerator 1. The block position information and the values ​​indicated in the blocks are used to control the combustion control device 120. Generates instructions.

[0050] Alternatively, the instruction unit 260 may be, for example, a crane control device 110 and a combustion control device 12. In this case, the inference map 4 generated by the information processing device 200 may be provided in both the 00 to the crane control device 110, and the crane control device 110 Generate instructions on which block on the map to put waste into incinerator 1, and then use crane 5. The combustion control device 120 then receives the inference map from the information processing device 200 and controls the operation of the 400 is received, and the crane control device 110 notifies the position of the block of waste thrown into the incinerator 1. The combustion control device 120 receives the placement information of the waste blocks fed into the waste incinerator 1. The block position information and the values ​​indicated in the blocks are used to generate control instructions for the incinerator 1.

[0051] FIG. 5 is a flow chart illustrating the operation of the waste treatment plant system 100 according to one embodiment. The rating is 500.

[0052] In step S510, the teacher data generating unit 220 uses the database 209 (see FIG. 2) The operation history of the waste treatment plant stored in the memory 204 (shown in FIG. 2) is read out. Then, from the operation history of the waste treatment plant loaded into the memory 204, and collecting process data corresponding to the image data of the waste in the waste pit 3. Based on the data, training data is generated.

[0053] In step S520, the model construction unit 230 creates an image data of the waste in the garbage pit 3. The training data corresponding to the image data generated in step S510 was used. Through supervised learning, the internal parameters of the functions of the model construction unit 230 are adjusted. Build a model that:

[0054] In step S530, the image acquisition unit 240 acquires the dust particles captured by the image capture device 6. Image data of the waste in the waste collection area 3 is acquired.

[0055] In step S540, the estimation unit 250 estimates the dust particles acquired by the image acquisition unit 240. The image data of the waste in the pit 3 is divided into one or more blocks. 0 inputs each of the divided image data into the model constructed in step S520. and obtain output data for each block.

[0056] In step S550, the estimation unit 250 calculates the block The value representing the quality of waste identified by the output data for each block is assigned to each block. An inference map 400 is generated.

[0057] In step S560, the instruction unit 260 executes the inference master generated in step S550. Based on the map 400, an operation command is sent to the crane control device 110 or the combustion control device 120. Give instructions.

[0058] The embodiments of the present invention have been described above. The present invention is not limited to the above and is intended to facilitate understanding of the invention. The present invention may be modified or improved without departing from the spirit thereof, and includes equivalents thereof. Of course, it is also possible to solve at least part of the above-mentioned problems. Alternatively, any combination of the embodiments and modifications may be used within the scope of achieving at least a part of the effects. Any combination of the elements described in the claims and the specification is possible. It is possible to combine or omit them. [Explanation of symbols]

[0059] 1... incinerator, 3... pit, 4... hopper, 5... crane, 6... imaging device, 100... waste disposal 1. A processing plant system, 110... a crane control device, 120... a combustion control device, 200... an information Processing device, 202...processor, 204...memory, 206...communication interface, 20 8...storage, 209...database, 220...teaching data generation unit, 230...model construction Construction section, 240...image acquisition section, 250...estimation section, 260...instruction section, 400...inference map

Claims

1. Using training data associated with images of waste stored in a garbage pit The model constructed through learning was then subjected to new imaging of waste stored in the waste pit. The proposed method involves inputting data from a new image and obtaining a value representing the quality of waste corresponding to the new image. Equipped with a fixed department, The estimation unit The new image of the waste is divided into a plurality of blocks, and the new image of the waste is divided into a plurality of blocks. and outputting a value representing the quality of the waste corresponding to the image as output data. Inference that associates information according to the output value representing the quality of waste with each of the blocks Generate a map, Whenever there is a change in the data of the new image, or at regular intervals, the inference map Update and / or record the An apparatus characterized in that

2. The estimation unit classifies the waste based on the magnitude of the value from the output value representing the quality of the waste. The label and a flag that determines whether or not the waste should be put into the incinerator, based on the magnitude of the value. and extracting at least one of the extracted labels and flags. to each of the blocks to generate the inference map.

2. The device of claim 1 .

3. When there are a plurality of pieces of output data corresponding to one block, the estimation unit performs the following for each block: The label and the like indicated in the block are selected from the plurality of output data corresponding to the block. and extracting at least one of the flags.

3. The device according to claim 2.

4. a model construction unit that constructs the model by learning using the training data 4. The apparatus of claim 1, further comprising:

5. The model construction unit acquires new images of the waste and corresponding process data. It has the function of periodically additionally learning or relearning using the 5. The device according to claim 4.

6. The device further comprises: Based on the inference map, instructions are given to a crane control device that controls the crane, or or an instruction to a combustion control device that controls an incinerator.

6. The device of claim 1, further comprising a display.

7. The instruction to the crane control device is to cause the crane to move the waste in the garbage pit. The instruction to the combustion control device is to operate the incinerator to burn the waste that has been put into the incinerator.

7. The device of claim 6, wherein the instructions are necessary to bake.

8. The training data is a data set obtained by capturing images of waste stored in the waste pit in a plurality of blocks. The image is divided into blocks, and the waste disposal plant corresponding to the image is acquired for each divided block. generated from the process data of the client, 8. Device according to any one of claims 1 to 7, characterized in that it

9. The process data is collected based on the operation history of the waste treatment plant. Based on data showing the characteristics of waste and the operation history of the waste treatment plant, workers classify the quality of the waste. and / or a similar label; 9. The device according to claim 8.

10. The value representing the quality of the waste is an index showing the combustibility of the waste.

10. The device according to any one of claims 9.

11. The training data associated with an image of the waste stored in the waste pit. The training data generation section generates 11. The apparatus of claim 1, further comprising:

12. A waste treatment plant system, comprising: a crane control device that controls the crane; a combustion control device that controls the incinerator; Using training data associated with images of waste stored in a garbage pit The model constructed through learning was then subjected to new imaging of waste stored in the waste pit. The new image data for each block is obtained by dividing the original image into multiple blocks. and inputting data into each of the waste quality data sets. an estimation unit that generates an inference map associated with the blocks; Based on the inference map, instructions to the crane control device or the combustion control an instruction unit that generates at least one of the instructions to the control device; Equipped with The estimation unit performs the following operation each time there is a change in the data of the new image, or at regular intervals: updating and / or recording the inference map; A system characterized by:

13. Using training data associated with images of waste stored in a garbage pit The model constructed through learning was then subjected to new imaging of waste stored in the waste pit. The proposed method involves inputting data from a new image and obtaining a value representing the quality of waste corresponding to the new image. a determining step, In the estimation step, The new image of the waste is divided into a plurality of blocks, and the new image of the waste is divided into a plurality of blocks. outputting a value representing the quality of the waste corresponding to the image; Inference that associates information according to the output value representing the quality of waste with each of the blocks Generate a map, The inference map is updated every time there is a change in the data of the new image, or at regular intervals. updated and / or recorded in A method characterized by:

14. The method of claim 13 is performed by a processor provided in a waste treatment plant. Program for.

15. A program for causing a computer to execute the method according to claim 13.

Images