JP2020032394A - Sewage treatment operation status evaluation device and sewage treatment operation status evaluation method - Google Patents

Abstract

To provide a sewage treatment operation status evaluation device and a sewage treatment operation status evaluation method, easy to be introduced, and possible to evaluate a sewage treatment operation status by accurately identifying microorganisms.SOLUTION: There is provided a sewage treatment operation status evaluation device that analyzes an input image obtained by imaging a microorganism living in a water treatment tank, and calculates which classification category the microorganism belongs to, for a sewage treatment facility where the microorganism purifies sewage stored in the water treatment tank, and evaluates an operation status of sewage treatment, the device including: a microbial feature learning unit 25 having machine learning equipment and configured to perform machine learning by inputting learning data including a learning image regarding the microorganism into the machine learning equipment, and create an identification parameter relating to identification of the microorganism; a microorganism identification unit 27 that calculates an identification category to which the microorganism belongs in the input image using the identification parameter; and an operation status evaluation unit 28 that evaluates an operation status of the sewage treatment based on an identification result in the microorganism identification unit 27.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a sewage treatment operation status evaluation device and a sewage treatment operation status evaluation method.

  BACKGROUND ART Conventionally, in a sewage treatment facility such as a sewage treatment facility, purification of sewage by activated sludge treatment has been widely performed. In activated sludge treatment, sewage is purified by exposing the sewage to air by aeration or the like to decompose pollutants into aerobic microorganisms in a water treatment tank that stores the sewage, proliferating the microorganisms, and separating them by sedimentation. Is done.

  In such a sewage treatment facility where activated sludge treatment is performed, generally, when sewage treatment is performed smoothly and when sewage treatment is performed in a deteriorated state instead of in a smooth state, the water treatment tank is It is known that the types of microorganisms that emerge and dominate are different. For example, when the state of the activated sludge is good, the genus Vorticella, the genus Epistylis, the shelled amoebae, the genus Arcella, and the rotifer, the genus Philodina, dominate. On the other hand, when the state of the activated sludge deteriorates, and in a transition period from a good state to a bad state, the flagellates Bodo, Monas, Amoeba, and Arcella appear.

Therefore, by identifying the microorganisms in the water treatment tank and associating the microorganisms with the state of the sewage treatment, it is possible to determine and evaluate whether or not the sewage treatment is being performed smoothly.
For example, activated sludge in a water treatment tank is sampled and observed with a microscope to confirm the presence of microorganisms that can serve as a standard for evaluating whether or not the sewage treatment is successful, as described above. However, since these operations are basically performed manually, it takes time. In addition, the worker needs advanced and specialized knowledge for identifying microorganisms, and skill cultivated through long working experience.

On the other hand, Patent Literature 1 discloses a biota diagnosis support system that identifies and counts microorganisms appearing in activated sludge in an activated sludge treatment process, and determines the treatment state of the process using a biota diagnosis device from the data. It has been disclosed. In this biota diagnosis support system, image data of treated water taken by a microscope camera or underwater camera is taken in by an input unit, and from this image data, an image recognition system automatically extracts microorganisms by image recognition processing using a model-based matching method. Identify and count.
Image models of microorganisms required for recognition are registered in a database in advance. Image recognition involves extracting the edges of a captured image of microorganisms, specifying the shape of the microorganisms using straight and arc components of the edge image, registering this shape in a database, and collating it with feature data obtained from the captured image of treated water. To identify the microorganism.

JP-A-8-197084

The biota diagnosis support system as disclosed in Patent Document 1 is based on a classic image processing technique such as edge extraction. That is, when creating an image model registered in advance in the database, it is necessary to effectively extract the characteristics of microorganisms. The task of extracting this feature requires advanced knowledge of image processing.
In addition, in order to enable the identification and identification of microorganisms by the biota diagnosis support system, it is necessary to intentionally extract characteristics of the microorganisms so that the characteristics of the microorganisms are remarkably expressed. Requires highly specialized knowledge.
In addition, it takes time and effort to extract the unique features of each microorganism to the extent that effective identification and identification by the biota diagnosis support system can be performed one by one for images of many microorganisms.
As described above, in the biota diagnosis support system as in Patent Literature 1, specialized knowledge is required, and it takes time and effort to prepare in advance.
Furthermore, since the shape of a microorganism changes due to movement, it is inherently difficult to accurately identify and identify the microorganism. This is precisely performed by the biota diagnosis support system based on the classical image processing technology as described above, for example, because the characteristics of the microorganisms are not represented by the photographed images used, and the identification of the microorganisms is performed. Is not easy because it may not be easy, and in fact, is not currently in practical use.

  The problem to be solved by the present invention is to provide a sewage treatment operation condition evaluation apparatus and a sewage treatment operation condition evaluation method that are easy to introduce and can evaluate the operation condition of sewage treatment by accurately identifying microorganisms. It is to be.

  The present invention employs the following means in order to solve the above problems. That is, the present invention is directed to a sewage treatment facility that purifies sewage stored in a water treatment tank with microorganisms, and analyzes an input image obtained by imaging the microorganisms living in the water treatment tank to determine which microorganisms are present. A sewage treatment operation status evaluation device that calculates whether a device belongs to an identification category and evaluates the operation status of sewage treatment, comprising a machine learning device, and inputting learning data including a learning image related to the microorganism to the machine learning device. Performing machine learning, a microorganism feature learning unit that generates an identification parameter related to the identification of the microorganism, a microorganism identification unit that calculates the identification classification to which the microorganism in the input image belongs using the identification parameter, An operation status evaluation unit that evaluates the operation status of the sewage treatment based on the identification result of the microorganism identification unit. To.

  Further, the present invention is directed to a sewage treatment facility for purifying sewage stored in a water treatment tank with microorganisms, and analyzes an input image obtained by imaging the microorganisms living in the water treatment tank to determine which microorganisms are present. Calculate whether it belongs to the identification section, evaluate the operating state of sewage treatment, is a sewage treatment operation state evaluation method, performing machine learning by inputting learning data including a learning image related to the microorganism to a machine learning device, Generating an identification parameter related to the identification of the microorganism, using the identification parameter, calculating the identification category to which the microorganism belongs in the input image, and evaluating an operation state of the wastewater treatment based on the identification result; A method for evaluating a processing operation state is provided.

  According to the present invention, it is possible to provide a sewage treatment operation condition evaluation apparatus and a sewage treatment operation condition evaluation method that are easy to introduce and can evaluate the operation condition of sewage treatment by accurately identifying microorganisms. .

It is a block diagram of a sewage treatment facility in a 1st embodiment of the present invention. It is a block diagram of a control device of the sewage treatment operation status evaluation device provided in the sewage treatment facility. It is a block diagram of a machine learning device which realizes a learning processing unit of the control device. It is a flowchart at the time of learning of the sewage treatment operation status evaluation method in the first embodiment. It is a flowchart at the time of sewage treatment operation status evaluation of the sewage treatment operation status evaluation method. It is a block diagram of the machine learning device which implement | achieves the learning processing part of the control apparatus of the sewage treatment operation condition evaluation apparatus in the modification of 1st Embodiment. It is a block diagram of a control device of a sewage treatment operation situation evaluation device in a 2nd embodiment of the present invention. It is a block diagram of the machine learning device which realizes the learning processing part of the control device of the sewage treatment operation status evaluation device in the second embodiment. It is a flowchart at the time of learning of the sewage treatment operation status evaluation method in the second embodiment. It is a flowchart at the time of sewage treatment operation status evaluation of the sewage treatment operation status evaluation method in the first embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
The sewage treatment operation status evaluation device in the present embodiment is directed to a sewage treatment facility that purifies sewage stored in a water treatment tank with microorganisms, and analyzes an input image of an image of microorganisms living in the water treatment tank. It is used to calculate which classification category a microorganism belongs to and evaluate the operating status of sewage treatment.It is equipped with a machine learning device, and inputs learning data including learning images related to microorganisms to the machine learning device to perform machine learning. A microbial feature learning unit that generates identification parameters related to the identification of microorganisms, a microbial identification unit that calculates the identification classification to which the microorganisms in the input image belong using the identification parameters, and a sewage based on the identification result in the microbial identification unit. An operation status evaluation unit that evaluates the operation status of the process.

[First Embodiment]
FIG. 1 is a block diagram of a wastewater treatment facility according to the first embodiment. The sewage treatment facility 1 purifies sewage stored in a water treatment tank with microorganisms.
The sewage treatment facility 1 includes a first settling tank 2, a biological treatment reaction tank (water treatment tank) 3, a reaction tank measuring device 4, a final settling tank 5, a diffuser plate 6, a blower 7, an aeration control valve 8, and a return sludge pump 9. , A sludge extraction pump 10, a sewage treatment operation status evaluation device 20, and first to fifth pipes L1, L2, L3, L4, L5.

  Sewage containing organic matter is introduced into the first settling basin 2. In the first sedimentation basin 2, small dirt, sand and the like in the introduced sewage are removed, and rough solid-liquid separation is performed. The sewage from which trash and the like have been removed is sent to the biological treatment reaction tank 3 via the first pipe L1.

In the biological treatment reaction tank 3, sewage is biologically treated and purified by microorganisms. At the time of purification, the sewage is exposed to air by aeration or the like, and aerobic microorganisms decompose the organic matter and multiply with the assimilation of the organic matter to form activated sludge. The water subjected to the activated sludge treatment is sent to the final sedimentation basin 5 via the second pipe L2.
Air is supplied to the lower part in the biological treatment reaction tank 3 from the blower 7 via the third pipe L3. An aeration adjusting valve 8 is provided in the third pipe L3 between the blower 7 and the biological treatment reaction tank 3. When the aeration control valve 8 is opened and closed, the amount of aeration changes, whereby the amount of dissolved oxygen in the biological treatment reaction tank 3 is adjusted, and the degree of progress of biological treatment by microorganisms is controlled. A diffuser plate 6 for improving the oxygen dissolving efficiency is provided in a portion of the biological treatment reaction tank 3 to which air is supplied by the third pipe L3.

In the final sedimentation basin 5, the activated sludge contained in the activated sludge-treated water sent from the biological treatment reaction tank 3 is settled to purify the sewage. The supernatant from which the activated sludge has been separated in the final sedimentation basin 5 is discharged out of the system as treated water.
Part of the sludge settled in the final settling tank 5 is returned to the biological treatment reaction tank 3 again through the fourth pipe L4 by the return sludge pump 9, and is reused for activated sludge treatment. The remaining sludge is discharged as surplus sludge by the sludge extraction pump 10 through the fifth pipe L5, sent to a sludge treatment facility or the like, and subjected to incineration after removing water.

The sewage treatment operation status evaluation device 20 according to the first embodiment is intended for the sewage treatment plant 1 that purifies sewage stored in the biological treatment reaction tank 3 with microorganisms as described above.
In the sewage treatment plant 1 as described above, when the sewage treatment is performed smoothly and when the sewage treatment is performed in a deteriorated state instead of smoothly, the sewage treatment appears in the biological treatment reaction tank 3 and becomes dominant. The type of microorganism used may be different. For example, when the state of the activated sludge is good, the genus Vorticella, the genus Epistylis, the shelled amoebae, the genus Arcella, and the rotifer, the genus Philodina, dominate. On the other hand, when the state of the activated sludge deteriorates, and in a transition period from a good state to a bad state, the flagellates Bodo, Monas, Amoeba, and Arcella appear.

The sewage treatment operation status evaluation device 20 according to the first embodiment identifies microorganisms in the biological treatment reaction tank 3 and associates the microorganisms with the state of the sewage treatment to determine whether the sewage treatment is being performed smoothly. Judge and evaluate. More specifically, the sewage treatment operation status evaluation device 20 analyzes an input image obtained by imaging the microorganisms living in the biological treatment reaction tank 3 to calculate which classification category the microorganism belongs to, and calculates the sewage treatment operation status. To evaluate.
That is, since the sewage treatment operation status evaluation device 20 analyzes the input image obtained by imaging the microorganism to identify the microorganism, the sewage treatment operation evaluation device 20 identifies the microorganism depending on its appearance characteristics.

Here, as an identification category in which the sewage treatment operation status evaluation device 20 identifies and classifies microorganisms, each of the microorganisms belonging to one class, for example, one "genus" in the biological classification has the same characteristics in the shape of these microorganisms. In addition, in a case where the activated sludge has a common effect, that is, a favorable or degraded effect, the “genus” may correspond to one identification category.
In addition, for example, in each of the microorganisms belonging to one "genus", the characteristics of the shape of these microorganisms are roughly classified into a plurality of types, and in each of the plurality of types, each microorganism belonging to the activated sludge condition. In the case where all the common types, that is, good or bad effects are provided, each of the roughly classified types may correspond to one identification section.
Alternatively, if there are a plurality of microorganisms having the same characteristic of the shape and the action on the activated sludge across the classification by the "genus", these may be collectively set as one identification category.

  In other words, the sewage treatment operation status evaluation device 20 according to the first embodiment has the same characteristics in shape, and the types of microorganisms are such that a plurality of microorganisms having a common action on activated sludge belong to one identification category. Is divided and classified into a plurality of identification categories, and at the time of evaluating the operation status of the sewage treatment, it is calculated to which identification category the microorganisms captured in the input image belong.

  Hereinafter, details of the sewage treatment operation status evaluation device 20 will be described. The sewage treatment operation status evaluation device 20 includes an imaging device 21 and a control device 22.

  In the first embodiment, the imaging device 21 is, for example, an underwater camera provided in the biological treatment reaction tank 3, and enlarges and shoots activated sludge in the biological treatment reaction tank 3. The imaging device 21 is not limited to this, and may be, for example, a system in which activated sludge in the imaging device 21 is automatically collected, enlarged by a microscope, and imaged.

The control device 22 is, for example, an information processing device such as a personal computer. FIG. 2 is a block diagram of the control device 22. The control device 22 includes a microorganism imaging unit 23, an image storage unit 24, a microorganism characteristic learning unit 25, a determination image acquisition unit 26, a microorganism identification unit 27, and an operation status evaluation unit 28. The microorganism feature learning unit 25 includes a learning image acquisition unit 30, a labeling unit 31, a learning processing unit 32, and a learning parameter storage unit 33.
Among these constituent elements of the control device 22, the microorganism imaging unit 23, the determination image acquisition unit 26, the microorganism identification unit 27, the driving situation evaluation unit 28, the learning image acquisition unit 30, the labeling unit 31, and the learning processing unit 32 are, for example, It may be software or a program executed by the CPU in the information processing apparatus. Further, the image storage unit 24 and the learning parameter storage unit 33 may be realized by a storage device such as a semiconductor memory or a magnetic disk provided inside and outside the information processing device.

As will be described later, the microorganism identification unit 27 calculates an identification classification to which the microorganism in the input image input to the control device 22 belongs, and based on this, the operation status evaluation unit 28 evaluates the operation status of the sewage treatment. . In order to effectively perform the identification of the microorganisms and the evaluation of the operation status of the sewage treatment, the microorganism characteristic learning unit 25 particularly includes a machine learning device as described later, and includes learning data including a learning image related to the microorganisms. Is input to a machine learning device to perform machine learning to generate identification parameters related to identification of microorganisms.
That is, the control device 22 roughly performs two operations, namely, learning of the classification of the microorganisms and evaluation of the driving situation. For the sake of simplicity, the following first describes each component of the control device 22 at the time of learning the identification classification, and then describes the behavior of each component at the time of evaluating the driving situation.

The microorganism imaging unit 23 controls the imaging device 21 to image the activated sludge in the biological treatment reaction tank 3 or automatically collected from the biological treatment reaction tank 3. The microbial imaging unit 23 further cuts out, for example, a part where an object is captured from the captured image, enlarges it to an appropriate size, and transmits the learning image to the image storage unit 24.
The microbial imaging unit 23 executes imaging and transmission to the image storage unit 24 until the number of learning images reaches the number required for appropriate learning by the machine learning device.

  The image storage unit 24 stores and stores a plurality of learning images transmitted from the microorganism imaging unit 23.

  The learning image acquisition unit 30 of the microorganism feature learning unit 25 acquires a learning image from the image storage unit 24 and transmits the learning image to the labeling unit 31.

The labeling unit 31 receives the learning image transmitted from the learning image acquisition unit 30.
The labeling unit 31 sequentially displays the learning images on a display device (not shown). The worker browses the learning image displayed on the display device, and if the learning image is a microorganism, associates a label corresponding to an identification section corresponding to the microorganism with the learning image.
The learning image associated with the label of the identification section becomes learning data of a machine learning device described later. That is, the machine learning device according to the first embodiment performs supervised learning by inputting, as teacher data, a learning image related to a microorganism and information on the identification classification of the microorganism corresponding to the learning image.
The labeling unit 31 transmits the learning data to the learning processing unit 32.

The learning processing unit 32 receives the learning data from the labeling unit 31.
The learning processing unit 32 includes a machine learning device 40 shown as a block diagram in FIG. The machine learning device 40 according to the first embodiment is for performing convolution deep learning, and is realized in particular by a convolutional neural network (Convolutional Neural Network). The machine learning device 40 generates a learned model 40A in which appropriate learning parameters have been learned, which is used as a program module that is a part of the artificial intelligence software.
The machine learning device 40 includes a convolution processing unit 41 and a full connection unit 42. The convolution processing unit 41 includes a first convolution layer 44, a second convolution layer 45, and a third convolution layer 46. The first convolution layer 44 includes an input layer 47 and an output layer 48.

The learning processing unit 32 inputs the learning image 43 to the machine learning device 40. In FIG. 3, the learning image 43 is shown as an image including three channels 43R, 43G, and 43B of R, G, and B.
The first convolution layer 44 includes a predetermined number of first filters (not shown). The machine learning device 40 positions each of the first filters on the learning image 43, and sets the pixel value of each pixel of the learning image 43 of the first filter corresponding to the pixel position in the first filter. The convolution filter processing is executed by adding the set weights and calculating the sum. Thereby, the pixel value of one pixel in the first convolution layer 44 is calculated. The machine learning device 40 calculates a plurality of pixel values by executing such a convolution filter process while moving the first filter on the learning image 43 at predetermined resolution intervals, and arranging the plurality of pixel values. Is generated. This image is referred to as a first feature map 44m because a feature corresponding to the first filter is extracted by the applied first filter.
The machine learning device 40 executes this processing for all the first filters, and generates the first feature maps 44m according to the number of the first filters.

  The first filter actually expresses emphasis or smoothing of the pixel value of the learning image 43 as a weight. In the first feature map 44m generated by executing the convolution filter processing using such a first filter, a light and shade pattern of an image such as an edge feature is effectively extracted. In addition, since the feature is extracted from the local region of the learning image 43 through the first filter, the position of the object existing in the learning image 43 is robust.

  In the first convolution layer 44, a pooling process is performed after the convolution filter process. More specifically, each of the first feature maps 44m is divided into a plurality of small areas having a size of, for example, 2 × 2, and the maximum value of the pixel values in the small area is calculated for each of the small areas. By setting the pixel value of the pixel, each 2 × 2 small area of each first feature map 44m is converted into a 1 × 1 pixel, and information is reduced. That is, in the first embodiment, the pooling processing is the maximum pooling processing. In the pooling process, as described above, since the maximum pixel value is selected from the local region of each first feature map 44m, it is possible to leave only appropriate features specific to the image. It can be done efficiently.

  The pooled first feature map 44m generated in the first convolution layer 44 becomes an input image of the second convolution layer 45.

In the second convolution layer 45, the convolution filter processing and the pooling processing are sequentially executed, as in the first convolution layer 44.
Like the first convolution layer 44, the second convolution layer 45 includes a predetermined number of second filters. The second convolution layer 45 executes a convolution filter process using the second filters, and further executes a pooling process. A predetermined number of pooled second feature maps 45m corresponding to the number of filters are generated.

Also in the third convolution layer 46, the convolution filter processing and the pooling processing are sequentially executed.
The third convolution layer 46 also includes a predetermined number of third filters, and performs a convolution filter process using these, and further executes a pooling process, so that a predetermined number corresponding to the number of the third filters is obtained. , A pooling-processed third feature map 46m is generated.
The weights of each of the first, second, and third filters are adjusted by machine learning.
The pixel value information of the pooled third feature map 46 m is input to the input layer 47.

  The input layer 47 includes a predetermined number of input nodes 47n. Each input node 47n is connected to all the pixel values of the pooled third feature map 46m, and has a connection weight between layers. In the input layer 47, a weighted sum is calculated for each pixel value information of the pooled third feature map 46m based on the connection weight, and an output function such as ReLU (Rectified Liner Unit) is calculated as a result. The applied value is stored in each input node 47n.

The machine learning device 40 according to the first embodiment learns to which of a plurality of identification categories the microorganisms in the input image belong, and the learned model 40A in which appropriate learning parameters have been learned by machine learning. Multi-class classification is performed. Therefore, assuming that the number of microorganism classification sections is N, the output layer 48 has a configuration including N output nodes 48n. This value N can be set, for example, for all types of microorganisms that may appear in one year in the sewage treatment facility 1 in which the sewage treatment operation status evaluation device 20 is installed.
Each output node 48n is connected to all the input nodes 47n of the input layer 47, and has a connection weight between the layers. In the output layer 48, a weighted sum is calculated for the input node 47n based on the connection weight, and a value obtained by applying an output function to the result is stored in the output node 48n.

In the machine learning device 40, the learning image 43 input to the first convolution layer 44 is processed in the convolution processing unit 41 and the full connection unit 42 as described above, and the processing result is stored in the output node 48n.
In the first embodiment, the value of the output node 48n corresponding to the identification category to which the microorganism in the input image belongs is close to a predetermined first determination value, for example, 1 and the value of the other output node 48n is predetermined. It is designed to be a second determination value, for example, a value close to 0.

In this case, the label corresponding to the learning image 43 input to the machine learning device 40, that is, the information of the identification classification to which the microorganism belongs is used as the correct value when the machine learning device 40 learns using the learning image 43. used. More specifically, it is assumed that the output node 48n corresponding to the classification to which the microorganism in the learning image 43 belongs should output the first determination value, and the other output nodes 48n should output the second determination value, respectively, The square error between these values and the value actually output from the machine learning device 40 is defined as a cost function.
Then, the weight value of each filter of the convolution processing unit 41, the value of each coupling weight of the total coupling unit 42, and the like are adjusted by an error back propagation method or the like so as to reduce this cost function. The learning device 40 performs machine learning.
The learning processing unit 32 uses the values of the weights of the filters of the convolution processing unit 41, the values of the connection weights of all the connection units 42, and the like at the time when the machine learning is completed as learning parameters, as a learning parameter storage unit 33. Send to

  The learning parameter storage unit 33 receives the learning parameters from the learning processing unit 32, accumulates and stores them.

As described above, the machine learning device 40 executes machine learning by generating the learned model 40A by reflecting the characteristics of the microorganism on the learning parameter. The learning parameters stored in the learning processing unit 32 are used as identification parameters relating to the identification of microorganisms in the evaluation of driving conditions described below. More specifically, in the evaluation of the driving situation, the learned model 40A having the learning parameters generated by the learning processing unit 32 is executed as a program on, for example, a CPU, so that the microorganisms in the image to be evaluated are obtained. Specifies the classification to which the belongs.
As described above, the microorganism feature learning unit 25 inputs the learning data to the machine learning device 40 to perform machine learning, and generates identification parameters related to identification of microorganisms.

  Next, the behavior of each component in the control device 22 at the time of evaluation of the driving situation will be described.

The microorganism imaging unit 23 controls the imaging device 21 to generate an image of the microorganism by imaging the activated sludge in the biological treatment reaction tank 3 or automatically collected from the biological treatment reaction tank 3, and inputs the microorganism. The image is transmitted to the image storage unit 24 as an image.
The microbial imaging unit 23 performs imaging and transmission to the image storage unit 24 until the number of these input images is equal to the number required for evaluation of the driving situation.

  The image storage unit 24 stores and stores a plurality of input images transmitted from the microorganism imaging unit 23.

  The determination image acquisition unit 26 acquires an input image from the image storage unit 24 and transmits the input image to the microorganism identification unit 27.

The microorganism identification unit 27 receives the input image transmitted from the determination image acquisition unit 26.
The microorganism identification unit 27 further acquires a learning parameter, that is, an identification parameter related to identification of a microorganism, from the learning parameter storage unit 33.
The microorganism identification unit 27 calculates an identification classification to which the microorganism in the input image belongs using the identification parameter. More specifically, the microorganism identifying unit 27 determines whether the weight value of each filter of the convolution processing unit 41, the value of each coupling weight of all the coupling units 42, and the like have been learned values as described above. That is, the learned model 40A is executed as a program on, for example, a CPU, and an input image is input to calculate the classification.

That is, as shown in FIG. 3, when the input image 49 is input to the first convolution layer 44 of the trained model 40A, the first convolution layer 44 sequentially executes the convolution filter processing and the pooling processing. Subsequently, the second and third convolution layers 45 and 46 execute a convolution filtering process and a pooling process, respectively, to generate a pooled third feature map 46m.
Further, by using the pooled third feature map 46m as an input, a process of calculating a weighted sum is executed in the all-coupling unit 42, and the processing result is finally stored in each output node 48n.

In each output node 48n, as described above, the value of the output node 48n corresponding to the classification to which the microorganism in the input image belongs is close to a predetermined first determination value, for example, 1 and the other output nodes 48n Is designed to be a predetermined second determination value, for example, a value close to 0. Therefore, for example, the classification corresponding to the output node 48n closest to the first determination value is identified and specified as the one to which the microorganism in the input image 49 belongs.
As described above, the microorganism identifying unit 27 inputs the input image 49 to the learned model 40A, and calculates the identification classification to which the microorganism in the input image 49 belongs.
The microorganism identification unit 27 calculates identification classifications corresponding to microorganisms for all the input images 49, and transmits all the calculated identification classifications to the operation status evaluation unit 28.

The operation status evaluation unit 28 receives from the microorganism identification unit 27 all calculated identification categories.
The operation status evaluation unit 28 evaluates the operation status of the sewage treatment based on the identification result of the microorganism identification unit 27. More specifically, all the received classifications are statistically analyzed, and microorganisms which become dominant when the state of activated sludge such as genus Vorticella, Epistylis, rusted amoebas Arcella, and rotifer Philodina are favorable are good. If the ratio of the identification categories corresponding to is equal to or greater than the predetermined first threshold, it is evaluated that the driving situation is good. Further, for example, identification classifications corresponding to microorganisms that appear when the state of activated sludge such as the flagellate Bodo genus, Monas genus, and Amoeba genus deteriorate, and in the transition period from a good state to a bad state, such as the genus Arcella Is greater than or equal to the second predetermined threshold, it is evaluated that the driving situation is deteriorating.
As described above, the operation status evaluation unit 28 evaluates the operation status of the sewage treatment based on the ratio of the microorganisms for each of the identification categories calculated by the microorganism identification unit 27 for the plurality of input images 49, that is, the appearance rate of the microorganisms. I do.

Next, a sewage treatment operation status evaluation method for evaluating the operation status by the sewage treatment operation status evaluation device 20 will be described with reference to FIGS. 1 to 3 and FIGS. 4 and 5. FIG. 4 is a flowchart at the time of learning in the sewage treatment operation status evaluation method. FIG. 5 is a flowchart of the sewage treatment operation status evaluation method when the sewage treatment operation status is evaluated.
This sewage treatment operation status evaluation method is intended for a sewage treatment plant 1 that purifies sewage stored in a water treatment tank with microorganisms. In order to calculate which classification class belongs to, and to evaluate the operation status of sewage treatment, learning data including a learning image relating to microorganisms is input to the machine learning device 40 to perform machine learning, and identification relating to identification of microorganisms is performed. A parameter is generated, a classification to which the microorganism in the input image belongs is calculated using the identification parameter, and an operation state of the sewage treatment is evaluated based on the identification result.
First, the operation at the time of learning the classification section will be described with reference to FIG.

The microorganism imaging unit 23 controls the imaging device 21 to image the activated sludge (Step S1). The microbial imaging unit 23 further cuts out, for example, a part where an object is captured from the captured image, enlarges it to an appropriate size, and transmits the learning image to the image storage unit 24.
The image storage unit 24 stores and stores a plurality of learning images transmitted from the microorganism imaging unit 23.

The learning image acquisition unit 30 of the microorganism feature learning unit 25 acquires a learning image from the image storage unit 24 and transmits the learning image to the labeling unit 31.
The labeling unit 31 receives the learning image transmitted from the learning image acquisition unit 30.
The labeling unit 31 sequentially displays the learning images on a display device (not shown). The operator browses the learning image displayed on the display device, and if the learning image is a microorganism, associates a label corresponding to an identification category corresponding to the microorganism with the learning image (step S3).
The labeling unit 31 transmits the learning data to the learning processing unit 32.

The learning processing unit 32 receives the learning data from the labeling unit 31.
The learning processing unit 32 inputs the learning image 43 to the machine learning device 40. In the machine learning device 40, the input learning image 43 is processed in the convolution processing unit 41 and the full connection unit 42, and the processing result is stored in the output node 48n.
By adjusting the weight value of each filter of the convolution processing unit 41, the value of each connection weight of the total connection unit 42, and the like by using the error back propagation method or the like to reduce the cost function for the processing result, Machine learning is performed by the machine learning device 40 (step S5).
The learning processing unit 32 uses the values of the weights of the filters of the convolution processing unit 41, the values of the connection weights of all the connection units 42, and the like at the time when the machine learning is completed as learning parameters, as a learning parameter storage unit 33. Send to

  The learning parameter storage unit 33 receives, accumulates, and stores the learning parameters from the learning processing unit 32 (Step S7).

  Next, the operation at the time of evaluating the driving situation will be described with reference to FIG.

The microorganism imaging unit 23 controls the imaging device 21 to image the activated sludge to generate an image of the microorganism, and transmits the image to the image storage unit 24 as an input image.
The image storage unit 24 stores and stores a plurality of input images transmitted from the microorganism imaging unit 23.
The determination image acquisition unit 26 acquires the input image 49 from the image storage unit 24 and transmits the input image 49 to the microorganism identification unit 27 (Step S11).

The microorganism identification unit 27 receives the input image 49 transmitted from the determination image acquisition unit 26.
The microorganism identification unit 27 further acquires a learning parameter, that is, an identification parameter related to identification of a microorganism, from the learning parameter storage unit 33.
The microorganism identification unit 27 calculates the identification classification to which the microorganism in the input image 49 belongs using the identification parameters (Step S13). More specifically, the microorganism identifying unit 27 determines whether the weight value of each filter of the convolution processing unit 41, the value of each coupling weight of all the coupling units 42, and the like have been learned values as described above. That is, the learned model 40A is executed as a program on, for example, a CPU, and an input image 49 is input to calculate an identification section.
The microorganism identification unit 27 calculates identification classifications corresponding to microorganisms for all the input images 49, and transmits all the calculated identification classifications to the operation status evaluation unit 28.

The operation status evaluation unit 28 receives from the microorganism identification unit 27 all calculated identification categories.
The operation status evaluation unit 28 evaluates the operation status of the wastewater treatment based on the identification result of the microorganism identification unit 27 (Step S15).

  Next, the effects of the above-described sewage treatment operation status evaluation apparatus and the sewage treatment operation status evaluation method will be described.

The sewage treatment operation status evaluation device 20 according to the first embodiment is intended for a sewage treatment plant 1 that purifies sewage stored in a biological treatment reaction tank (water treatment tank) 3 with microorganisms. A sewage treatment operation status evaluation apparatus 20 for analyzing an input image 49 obtained by capturing an image of a microorganism that inhabits the soil, calculating which classification the microorganism belongs to, and evaluating an operation status of the sewage treatment. The learning data including the learning image 43 related to the microorganisms is input to the machine learning device 40 to perform machine learning, and the microorganism feature learning unit 25 that generates the identification parameter related to the identification of the microorganism. A microorganism identification unit 27 that calculates an identification category to which the microorganisms in 49 belong, and an operation status that evaluates an operation status of sewage treatment based on the identification result in the microorganism identification unit 27 It includes a valence section 28.
According to the configuration as described above, the microbial feature learning unit 25 inputs the learning data including the learning image 43 on the microbe to the machine learning device 40 and performs the machine learning. It is possible to automatically generate an identification parameter that is extracted and reflects this. For this reason, when introducing the sewage treatment operation situation evaluation device 20, a high degree of image processing knowledge is not particularly required.
Further, since there is no need for the worker to intentionally extract the characteristics so that the characteristics unique to the microorganism are remarkably expressed, highly specialized knowledge on the microorganism is not particularly required. In the first embodiment, it is necessary for the operator to associate the label with the microorganism, but it is not required to have a particularly high level of knowledge required for extracting the characteristics of the microorganism.
Therefore, introduction of the sewage treatment operation status evaluation device 20 is easy.
In addition, the accuracy of the identification parameter can be easily improved by preparing a large amount of learning data to be input to the machine learning device 40.
Further, when calculating the identification classification to which the microorganism belongs, the microorganism identification unit 27 uses the identification parameter generated as described above. The identification parameter reflects a feature of the machine learning device 40 that abstracts the appearance of the microorganism. For this reason, when calculating the classification to which the microorganism belongs, it is possible to suppress the influence of the imaged angle, size, color, weather, and time-dependent lightness of the microorganism, and the like.
Therefore, it is possible to increase the calculation accuracy of the identification section to which the microorganism belongs, that is, the accuracy of evaluating the operation state of the wastewater treatment.

The machine learning device 40 performs convolution deep learning. In particular, in the first embodiment, the machine learning device 40 is realized by a convolutional neural network.
According to the above configuration, the sewage treatment operation status evaluation device 20 can be appropriately realized.

In addition to the learning image 43, the machine learning device 40 inputs information on the identification classification of the microorganism corresponding to the learning image 43 as teacher data and reflects the characteristics of the microorganism on the learning parameter to generate the learned model 40A. To perform machine learning, and the identification parameters are learning parameters.
Further, the microorganism identification unit 27 inputs the input image 49 to the learned model 40A, and calculates an identification classification to which the microorganism in the input image 49 belongs.
For example, in a conventional system based on a classic image processing technique, the brightness, image quality, and the like of a captured image input to the system greatly depend on the environment of the sewage treatment facility 1. For this reason, in evaluating the driving situation, various parameters in the image processing may need to be appropriately adjusted according to the imaging environment.
On the other hand, according to the above-described configuration, the machine learning device 40 executes the machine learning to generate the learned model 40A, and the microorganism identifying unit 27 determines the learning parameter, that is, the learned model 40A in which the identification parameter is reflected. , An input image 49 is input, and the classification to which the microorganism in the input image 49 belongs is calculated. That is, since the input image 49 used in the evaluation of the driving condition is processed by the learned model 40A in which the characteristics of the microorganisms are reflected, the characteristics are efficiently extracted, so that the influence of the imaging environment can be reduced. Can be. Therefore, the operation of the sewage treatment operation status evaluation device 20 is facilitated.

Further, the machine learning device 40 learns to which of the plurality of identification sections the microorganism in the learning image 43 belongs.
According to the above configuration, since the classification to which the microorganism belongs can be calculated by only one machine learning device 40, the number of machine learning devices 40 provided in the sewage treatment operation status evaluation device 20 can be reduced. . That is, since the number of machine learning devices 40 to be machine-learned is reduced, it is easy to introduce the sewage treatment operation status evaluation device 20.

In addition, the operation status evaluation unit 28 evaluates the operation status of the sewage treatment based on the appearance rates of microorganisms for each of the identification categories calculated by the microorganism identification unit 27 for the plurality of input images 49.
According to the above configuration, the operating state of the sewage treatment can be evaluated with high accuracy.

In particular, the sewage treatment operation status evaluation device 20 as described above is effective when there are many types of microorganisms. When the types of microorganisms increase, even if there are similar microorganisms, there is a possibility that the influence on the driving situation will be different.For example, in a system based on the classic image processing technology of the related art, the microorganisms may be mistaken. As a result, the driving situation may be erroneously evaluated. In addition, in the conventional system, since a worker must extract characteristics according to the type of microorganism, the labor required for introduction increases in proportion to the type of microorganism.
On the other hand, in the sewage treatment operation status evaluation device 20, even when the types of microorganisms increase, the machine learning device 40 learns to accurately capture the characteristics of the microorganisms and to identify and calculate each identification classification. If the erroneous evaluation of the driving situation can be suppressed, and if the learning data is created, the feature extraction itself is automatically performed, so that the increase in the introduction cost is minute. Thus, each of the above-mentioned effects already described for the sewage treatment operation status evaluation device 20 can be more remarkable when there are many types of microorganisms as compared with the conventional system.

[Modification of First Embodiment]
Next, a modified example of the sewage treatment operation status evaluation apparatus and the sewage treatment operation status evaluation method shown as the first embodiment will be described with reference to FIG. FIG. 6 is a block diagram of the machine learning device 50 provided in the learning processing unit of the sewage treatment operation status evaluation device in the present modification. The sewage treatment operation status evaluation device of the present modification is different from the sewage treatment operation status evaluation device 20 of the first embodiment in the configuration of the machine learning device 50.

In the first embodiment, the learning processing unit 32 includes only one machine learning device 40 that learns the N class classification when the number of microorganism classification sections is N. The machine learning device 50 included in the learning processing unit according to the present modification includes N machine learning devices 50-1 to 50-N.
Each of the machine learning devices 50-1 to 50-N corresponds to each of the identification sections of the microorganism, and learns whether or not the microorganisms in the input image belong to the corresponding identification section. That is, in this modified example, each learned model in which appropriate learning parameters have been learned by machine learning performs two-class classification.

More specifically, each output layer 58 of the machine learning devices 50-1 to 50-N includes two output nodes 58a and 58b. Each of the machine learning devices 50-1 to 50-N has a value of one output node 58a close to a predetermined first determination value, for example, 1 when there is no microorganism of the corresponding identification classification in the input image, The value of the output node 58b is close to the second determination value when the value of the other output node 58b is close to a predetermined second determination value, for example, 0, and there is a microorganism. Is designed to be a value close to the first determination value.
For each of the machine learning devices 50-1 to 50-N, a learning image related to the microorganism and information on whether or not the microorganism in the learning image belongs to the identification section corresponding to the machine learning device is included in the teacher data. And supervised learning is performed.
The learning parameter storage unit stores the weight of each filter of the convolution processing unit 51 and the weight of each connection weight of the total connection unit 52 for each of the plurality of machine learning devices 50-1 to 50-N that have been machine-learned in this manner. Values and the like are accumulated and stored as learning parameters.

  The microorganism identification unit inputs each of the input images received from the determination image acquisition unit to all the learned models corresponding to each of the machine learning devices 50-1 to 50-N, and performs a corresponding operation based on each learned model. The identification section to which the microorganism in the input image belongs is calculated by summing up the determination results of the presence / absence of the microorganism belonging to the identification section.

It is needless to say that the present modified example has the same effects as those of the first embodiment already described.
In particular, in the present modification, the corresponding machine learning devices 50-1 to 50-N are provided for each classification of the microorganisms, and each of the machine learning devices 50-1 to 50-N is configured such that the microorganisms in the learning image are It learns whether it belongs to the corresponding identification section.
According to the above configuration, each of the machine learning devices 50-1 to 50-N is machine-learned specifically for the corresponding identification section, so that the identification section of the microorganism can be calculated more accurately. .
Further, for example, when an image obtained by capturing a plurality of microorganisms is input as an input image, in the sewage treatment operation status evaluation device 20 of the first embodiment, the learned model 40A determines the identification classification to which the microorganism belongs. As a result, a situation may arise in which the plurality of output nodes 48n corresponding to each of the identification sections to which the plurality of microorganisms belong output, for example, an intermediate value between the first determination value and the second determination value. On the other hand, in the configuration of the present modified example, each learned model operates to detect only the microorganisms belonging to the corresponding identification section. The output node 58b of the learned model 50-1 to 50-N corresponding to the section outputs a value close to the first determination value. Therefore, even when a plurality of microorganisms are captured in the input image, the identification section corresponding to each microorganism can be accurately calculated and specified.

[Second embodiment]
Next, a wastewater treatment operation status evaluation device 60 according to the second embodiment will be described. In the second embodiment, the same components as those in the first embodiment are denoted by the same reference numerals in the drawings, and description thereof will be omitted.
The sewage treatment operation status evaluation device 60 is used in the sewage treatment facility 1 described with reference to FIG. 1, similarly to the sewage treatment operation status evaluation device 20 in the first embodiment.
The sewage treatment operation status evaluation device 60 includes an imaging device 21 and a control device 62.
The imaging device 21 is the same as in the first embodiment.

The control device 62 is an information processing device such as a personal computer. FIG. 7 is a block diagram of the control device 62. The control device 62 includes a microorganism imaging unit 23, an image storage unit 24, a microorganism characteristic learning unit 65, a determination image acquisition unit 26, a characteristic amount extraction unit 68, a characteristic amount accumulation unit 69, a determination characteristic amount acquisition unit 70, and a microorganism identification unit 71. , And an operation status evaluation unit 28. The microorganism feature learning unit 65 includes a machine learning unit 66 and a clustering unit 67. The machine learning unit 66 includes a learning image acquisition unit 30, a learning processing unit 81, and a learning parameter storage unit 82. The clustering unit 67 includes a learning feature amount acquisition unit 83, a clustering execution unit 84, a clustering parameter storage unit 85, a result presentation unit 86, and an association storage unit 88.
Among the components of the control device 22, the microorganism imaging unit 23, the determination image acquisition unit 26, the feature amount extraction unit 68, the determination feature amount acquisition unit 70, the microorganism identification unit 71, the operation status evaluation unit 28, the learning image acquisition unit 30 The learning processing unit 81, the learning feature amount acquisition unit 83, the clustering execution unit 84, and the result presentation unit 86 may be software or programs executed by, for example, a CPU in the information processing apparatus. Further, the image storage unit 24, the feature amount storage unit 69, the learning parameter storage unit 82, the clustering parameter storage unit 85, and the association storage unit 88 store storages such as a semiconductor memory and a magnetic disk provided inside and outside the information processing apparatus. It may be realized by an apparatus.

  As in the first embodiment, the control device 62 roughly performs two types of operations: learning of the classification of microorganisms and evaluation of the driving situation. For the sake of simplicity, the following describes the components of the control device 62 at the time of learning the identification classification, and then describes the behavior of the components at the time of evaluating the driving situation.

The microorganism imaging unit 23 controls the imaging device 21 to image the activated sludge. The microbial imaging unit 23 further cuts out, for example, a part where an object is captured from the captured image, enlarges it to an appropriate size, and transmits the learning image to the image storage unit 24.
The microbial imaging unit 23 executes imaging and transmission to the image storage unit 24 until the number of learning images reaches the number required for appropriate learning by the machine learning device.

  The image storage unit 24 stores and stores a plurality of learning images transmitted from the microorganism imaging unit 23.

The learning image acquiring unit 30 of the machine learning unit 66 of the microbial feature learning unit 65 acquires a learning image from the image storage unit 24 and transmits the acquired learning image to the learning processing unit 81 as learning data.
The learning image acquisition unit 30 also transmits the learning image to a feature amount extraction unit 68 described later.

The learning processing unit 81 receives the learning data from the learning image acquisition unit 30.
The learning processing unit 81 includes a machine learning device 90 shown as a block diagram in FIG. The machine learning device 90 according to the second embodiment performs convolutional deep learning, and is realized in particular by a convolutional autoencoder (Convolutional Autoencoder). The machine learning device 90 is for generating a learned model in which appropriate learning parameters have been learned, which is used as a program module that is a part of the artificial intelligence software.
The machine learning device 90 includes a convolution processing unit 91, a fully connected layer 92, and a deconvolution processing unit 93. The convolution processing section 91 includes first to third convolution layers 91a, 91b, and 91c. The total coupling layer 92 includes first to third total coupling layers 92a, 92b, and 92c. The deconvolution processing unit 93 includes first to third deconvolution layers 93a, 93b, 93c.

The learning processing unit 81 inputs the learning image 43 to the machine learning device 90. The first to third convolution layers 91a, 91b, and 91c of the convolution processing unit 91 generate a feature map 91m to be input to the fully connected layer 92 by the same processing as that of the convolution processing unit 41 in the first embodiment.
The third convolutional layer 91c is connected to the first fully coupled layer 92a of the fully coupled layer 92 having a predetermined number of nodes 921. The value of the feature map 91m is stored in each node 921 of the first fully connected layer 92a, as in the first embodiment.

  The second fully coupled layer 92b also has a predetermined number of nodes 92m, like the first fully coupled layer 92a. The number of nodes 92m of the second fully coupled layer 92b is smaller than the number of nodes 921 of the first fully coupled layer 92a. Each node 92m of the second fully coupled layer 92b is coupled to all nodes 921 of the first fully coupled layer 92a, and has a coupling load between the layers. In the second fully coupled layer 92b, based on this coupling weight, a weighted sum is calculated for the node 921 of the first fully coupled layer 92a, and the value obtained by applying the output function to the result is the second fully coupled layer 92b. It is stored in each node 92m of the node 92b.

  The third fully coupled layer 92c has the same number of nodes 92n as the first fully coupled layer 92a. Each node 92n of the third fully coupled layer 92c is coupled to all nodes 92m of the second fully coupled layer 92b, and has a coupling load between the layers. In the third fully coupled layer 92c, based on the coupling load, a weighted sum is calculated for the node 92m of the second fully coupled layer 92b, and a value obtained by applying an output function to the result is the third fully coupled layer 92c. It is stored in each node 92n of the node 92c.

The value of each node 92n of the third fully connected layer 92c is input to the deconvolution processing unit 93. In the deconvolution processing unit 93, processing symmetrical to that of the convolution processing unit 91 is executed. That is, although the learning image 43 is compressed by the convolution processing unit 91 to a low dimension before reaching the fully connected layer 92, the deconvolution processing unit 93 operates so as to be restored from the low-dimensional compression state. . More specifically, a first deconvolution layer 93a that performs processing substantially symmetric with the third convolution layer 91c, a second deconvolution layer 93b that performs processing that is substantially symmetric with the second convolution layer 91b, and a first convolution layer An image 94 having the same resolution as the learning image 43 is generated by sequentially passing through the third deconvolution layer 93c that performs a process substantially symmetric to that of the learning image 43.
In the machine learning device 90, machine learning is performed so that the generated image 94 becomes a reproduced image 94 that reproduces the learning image 43. That is, the machine learning device 90 according to the second embodiment performs learning using only the learning image 43 relating to the microorganisms as learning data without particularly requiring information generated by a worker's label association work.

The learning of the machine learning device 90 is executed to generate a learned model used in calculating the classification to which the microorganism belongs, as in the first embodiment, but is different from the first embodiment. Differently, not all parameters set in the learning of the machine learning device 90 are used as the learned model.
As already described, the machine learning device 90 is a convolutional auto-encoder, and the learning image 43 is compressed to a low dimension in the convolution processing unit 91 and is input to the first fully connected layer 92a of the fully connected layer 92, and the second After being maximally compressed in the fully connected layer 92b, it is restored to a reproduced image 94. That is, the second fully connected layer 92b is a portion where the features of the input image are in the most compressed state, and the value of the node 92m is a feature amount obtained by abstractly condensing the features of the microorganisms in the image. Can handle.

The second embodiment conceptually utilizes this characteristic of the convolutional auto-encoder, inputs each of the learning images 43 to the machine learning device 90 that has completed learning, and generates a second image corresponding to each learning image 43. (2) The value of each node 92m of the fully connected layer 92b is stored as a feature value. In this state, the trained model is executed on the image input during the evaluation of the driving situation to derive the value of each node 92m of the second fully connected layer 92b, that is, the feature amount, for the image, and save this in advance. The classification of the microorganisms in the input image is calculated by comparing with the feature amount of the learning image 43 thus obtained.
Therefore, in the second embodiment, the portion from the first convolution layer 91a to the second fully connected layer 92b, which is surrounded by the two-dot chain line in FIG. 8, corresponds to the learned model 95. Therefore, the value of each parameter in this portion, that is, the value of the weight of each filter of the convolution processing unit 91, the value of each coupling load of the total coupling layer 92 up to the second total coupling layer 92b, and the like are used in the second embodiment. Is the learning parameter in.
The learning processing unit 81 transmits the learning parameters at the time when the machine learning ends, to the learning parameter storage unit 82.

  As described above, the machine learning unit 66 of the microorganism feature learning unit 65 inputs the learning image 43 on the microorganism as learning data to the machine learning device 90 to perform machine learning. The machine learning device 90 executes the machine learning by inputting the learning data and reflecting the characteristics of the microorganism on the learning parameter to generate the learned model 95.

  The learning parameter storage unit 82 receives, accumulates, and stores the learning parameters from the learning processing unit 81.

The feature amount extraction unit 68 receives each learning image 43 from the learning image acquisition unit 30.
In addition, the feature amount extraction unit 68 acquires a learning parameter from the learning parameter storage unit 82.
The feature amount extraction unit 68 learns the learning parameters, that is, the values of the weights of the filters of the convolution processing unit 91 and the values of the connection weights of the total connection layer 92 up to the second full connection layer 92b, as described above. The learned model 95 having the value is executed as a program on a CPU, for example, and the value of each node 92m of the second fully connected layer 92b, that is, the feature amount corresponding to each learning image 43 is calculated.
As described above, the feature amount extraction unit 68 inputs the learning image 43 again to the learned model 95, and extracts the feature of the microorganism in the learning image 43 as the feature amount.
The feature amount extraction unit 68 transmits a combination of the learning image 43 and the calculated feature amount to the feature amount accumulation unit 69 for each learning image 43.

The feature storage unit 69 receives the combination of the learning image 43 and the feature from the feature extraction unit 68, and stores and stores the combination.
The combination of the learning image 43 and the feature amount stored in the feature amount accumulation unit 69 is used in the clustering unit 67 described below. The clustering unit 67 divides a plurality of feature amounts corresponding to each of the learning images 43 into a plurality of sets corresponding to each identification section by unsupervised clustering.

  The learning feature amount acquiring unit 83 of the clustering unit 67 of the microbial feature learning unit 65 acquires all combinations of the learning image 43 and the feature amount from the feature amount accumulating unit 69 and transmits the combination to the clustering execution unit 84.

The clustering execution unit 84 receives from the learning feature amount acquisition unit 83 all combinations of the learning image 43 and the feature amount.
The clustering execution unit 84 executes clustering on all feature amounts. As a clustering method, a method assuming a statistical model such as a mixture normal distribution, a machine learning method such as SVM (Support Vector Machine), or the like can be applied.
As a result of the clustering, each feature amount is classified into a plurality of sets, and each feature amount is assigned a cluster ID associated with the corresponding set.
The clustering execution unit 84 transmits to the clustering parameter storage unit 85 the clustering parameters set as a result of the classification in the clustering unit and all combinations of the learning image 43, the feature amount, and the cluster ID.

  The clustering parameter storage unit 85 receives, accumulates, and stores the clustering parameters and all combinations of the learning image 43, the feature amount, and the cluster ID from the clustering execution unit 84.

The result presentation unit 86 acquires the combination of the learning image 43, the feature amount, and the cluster ID from the clustering parameter storage unit 85.
The result presenting unit 86 presents the learning image 43 representing each set and the corresponding cluster ID to the worker based on the acquired information by using a display device (not shown) or the like.
In the above description, as the learning image 43 representing each set, for example, in the case of clustering assuming a mixture normal distribution, an image having a parameter closest to the (average, variance) of each mixture component is used in the case of classification by SVM. In, an image located farthest from the class boundary in the parameter space can be considered.
Based on this, the operator assigns an identification section of the corresponding microorganism, for example, a label indicating the type, to each cluster ID using an input device or the like (not shown).
The result presentation unit 86 transmits all combinations of the cluster ID and the assigned label to the association storage unit 88.

  The association storage unit 88 receives, stores, and stores all combinations of the cluster ID and the label from the result presenting unit 86.

  Next, the behavior of each component at the time of evaluation of the driving situation in the control device 62 will be described.

The microorganism imaging unit 23 controls the imaging device 21 to generate an image of the microorganism, and transmits the image to the image storage unit 24 as an input image.
The microbial imaging unit 23 performs imaging and transmission to the image storage unit 24 until the number of these input images is equal to the number required for evaluation of the driving situation.

  The image storage unit 24 stores and stores a plurality of input images transmitted from the microorganism imaging unit 23.

  The determination image acquisition unit 26 acquires an input image from the image storage unit 24 and transmits the input image to the feature amount extraction unit 68.

The feature amount extraction unit 68 receives the input image transmitted from the determination image acquisition unit 26.
In addition, the feature amount extraction unit 68 acquires a learning parameter from the learning parameter storage unit 82.
The feature amount extraction unit 68 executes the learned model 95 in which the learning parameters are reflected, for example, as a program on a CPU, and obtains the feature amount corresponding to the input image 49, which is the value of each node 92m of the second fully connected layer 92b. Is calculated. That is, the feature amount extraction unit 68 inputs the input image 49 to the trained model 95 and extracts a feature of the microorganism in the input image 49 as a feature amount.
The feature value extraction unit 68 transmits a feature value to the feature value storage unit 69 for each input image 49.

  The feature amount accumulating unit 69 receives a feature amount corresponding to each input image 49 from the feature amount extracting unit 68, accumulates and stores the feature amount.

  The determination feature amount acquiring unit 70 acquires a feature amount corresponding to each input image 49 from the feature amount accumulating unit 69, and transmits the acquired feature amount to the microorganism identifying unit 71.

The microorganism identification unit 71 receives all the feature values transmitted from the determination feature value acquisition unit 70.
The microorganism identification unit 71 acquires, from the clustering parameter storage unit 85, the clustering parameters set as a result of the classification in the clustering unit.
The microorganism identification unit 71 further acquires all combinations of the cluster ID and the label from the association storage unit 88.
The microorganism identification unit 71 calculates a cluster ID for each feature generated from the input image 49 using the clustering parameters. The microorganism identification unit 71 calculates, for each feature amount, a label corresponding to the cluster ID, that is, an identification category to which the microorganism belongs.
As described above, the microorganism identification unit 71 calculates the classification to which the microorganism belongs by calculating which of the plurality of sets the characteristic amount of the microorganism in the input image 49 is based on the clustering parameter. .

As described above, in the second embodiment, the microorganism identification unit 71 uses the clustering parameters generated by the microorganism characteristic learning unit 65 and stored in the clustering parameter storage unit 85 as identification parameters related to identification of microorganisms. , The classification to which the microorganism in the input image 49 belongs is calculated.
As described above, the identification parameter, that is, the clustering parameter in the second embodiment is generated by clustering the feature amount of each learning image generated based on the learning parameter in the learned machine learning device 90. ing.

  The microorganism identification unit 71 transmits all the calculated identification classifications to the operation status evaluation unit 28.

The operation status evaluation unit 28 receives from the microorganism identification unit 71 all calculated identification categories.
The operation status evaluation unit 28 evaluates the operation status of the wastewater treatment based on the identification result obtained by the microorganism identification unit 71. More specifically, all the received classifications are statistically analyzed, and microorganisms which become dominant when the state of activated sludge such as genus Vorticella, Epistylis, rusted amoebas Arcella, and rotifer Philodina are favorable are good. If the ratio of the identification categories corresponding to is equal to or greater than the predetermined first threshold, it is evaluated that the driving situation is good. Further, for example, identification classifications corresponding to microorganisms that appear when the state of activated sludge such as the flagellate Bodo genus, Monas genus, and Amoeba genus deteriorate, and in the transition period from a good state to a bad state, such as the genus Arcella Is greater than or equal to the second predetermined threshold, it is evaluated that the driving situation is deteriorating.
As described above, the operation status evaluation unit 28 evaluates the operation status of the sewage treatment based on the ratio of the microorganisms for each of the identification categories calculated by the microorganism identification unit 71 for the plurality of input images 49, that is, the appearance rate of the microorganisms. I do.

Next, a sewage treatment operation status evaluation method for evaluating the operation status by the sewage treatment operation status evaluation device 60 will be described with reference to FIGS. 1, 7, 8, 9, and 10. FIG. 9 is a flowchart at the time of learning in the sewage treatment operation status evaluation method. FIG. 10 is a flowchart at the time of sewage treatment operation status evaluation of the sewage treatment operation status evaluation method.
First, the operation at the time of learning the identification section will be described with reference to FIG.

The microorganism imaging unit 23 controls the imaging device 21 to image the activated sludge (Step S21). The microbial imaging unit 23 further cuts out, for example, a part where an object is captured from the captured image, enlarges it to an appropriate size, and transmits the learning image to the image storage unit 24.
The image storage unit 24 stores and stores a plurality of learning images transmitted from the microorganism imaging unit 23.

The learning image acquiring unit 30 of the machine learning unit 66 of the microbial feature learning unit 65 acquires a learning image from the image storage unit 24 and transmits the acquired learning image to the learning processing unit 81 as learning data.
The learning image acquisition unit 30 also transmits the learning image to the feature amount extraction unit 68.

The learning processing unit 81 receives the learning data from the learning image acquisition unit 30.
The learning processing unit 81 inputs the learning image 43 to the machine learning device 90, and performs machine learning on the machine learning device 90.
The learning processing unit 81 transmits the learning parameters at the time when the machine learning is completed to the learning parameter storage unit 82.

  The learning parameter storage unit 82 receives, accumulates, and stores the learning parameters from the learning processing unit 81.

The feature amount extraction unit 68 receives each learning image 43 from the learning image acquisition unit 30.
In addition, the feature amount extraction unit 68 acquires a learning parameter from the learning parameter storage unit 82.
The feature amount extraction unit 68 learns the learning parameters, that is, the values of the weights of the filters of the convolution processing unit 91 and the values of the connection weights of the total connection layer 92 up to the second full connection layer 92b, as described above. The learned model 95 having the value is executed as a program on a CPU, for example, and the value of each node 92m of the second fully connected layer 92b, that is, the feature amount corresponding to each learning image 43 is calculated.
As described above, the feature amount extraction unit 68 inputs the learning image 43 to the learned model 95 and extracts the feature of the microorganism in the learning image 43 as the feature amount (Step S23).
The feature amount extraction unit 68 transmits a combination of the learning image 43 and the calculated feature amount to the feature amount accumulation unit 69 for each learning image 43.

  The feature storage unit 69 receives the combination of the learning image 43 and the feature from the feature extraction unit 68, and stores and stores the combination.

  The learning feature amount acquiring unit 83 of the clustering unit 67 of the microbial feature learning unit 65 acquires all combinations of the learning image 43 and the feature amount from the feature amount accumulating unit 69 and transmits the combination to the clustering execution unit 84.

The clustering execution unit 84 receives from the learning feature amount acquisition unit 83 all combinations of the learning image 43 and the feature amount.
The clustering execution unit 84 executes clustering on all feature amounts (step S25).
As a result of the clustering, each feature amount is classified into a plurality of sets, and each feature amount is assigned a cluster ID associated with the corresponding set.
The clustering execution unit 84 transmits to the clustering parameter storage unit 85 the clustering parameters set as a result of the classification in the clustering unit and all combinations of the learning image 43, the feature amount, and the cluster ID.

  The clustering parameter storage unit 85 receives, accumulates, and stores the clustering parameters and all combinations of the learning image 43, the feature amount, and the cluster ID from the clustering execution unit 84 (step S27).

The result presentation unit 86 acquires the combination of the learning image 43, the feature amount, and the cluster ID from the clustering parameter storage unit 85.
The result presenting unit 86 presents the learning image 43 representing each set and the corresponding cluster ID to the worker based on the acquired information by using a display device (not shown) or the like.
The operator assigns an identification classification of the corresponding microorganism, for example, a label indicating the type to each cluster ID using an input device or the like (not shown) (step S29).
The result presentation unit 86 transmits all combinations of the cluster ID and the assigned label to the association storage unit 88.

  The association storage unit 88 receives, stores, and stores all combinations of the cluster ID and the label from the result presenting unit 86.

  Next, the operation at the time of evaluating the driving situation will be described with reference to FIG.

The microorganism imaging unit 23 controls the imaging device 21 to generate an image of the microorganism, and transmits the image to the image storage unit 24 as an input image.
The image storage unit 24 stores and stores a plurality of input images transmitted from the microorganism imaging unit 23.
The determination image acquisition unit 26 acquires an input image from the image storage unit 24 and transmits the input image to the feature amount extraction unit 68.

The feature amount extraction unit 68 receives the input image transmitted from the determination image acquisition unit 26.
In addition, the feature amount extraction unit 68 acquires a learning parameter from the learning parameter storage unit 82.
The feature amount extraction unit 68 executes the learned model 95 in which the learning parameters are reflected, for example, as a program on a CPU, and obtains the feature amount corresponding to the input image 49, which is the value of each node 92m of the second fully connected layer 92b. Is calculated. That is, the feature amount extraction unit 68 inputs the input image 49 to the learned model 95, and extracts a feature of the microorganism in the input image 49 as a feature amount (step S31).
The feature value extraction unit 68 transmits a feature value to the feature value storage unit 69 for each input image 49.

The feature amount accumulating unit 69 receives a feature amount corresponding to each input image 49 from the feature amount extracting unit 68, accumulates and stores the feature amount.
The determination feature amount acquiring unit 70 acquires a feature amount corresponding to each input image 49 from the feature amount accumulating unit 69, and transmits the acquired feature amount to the microorganism identifying unit 71.

The microorganism identification unit 71 receives all the feature values transmitted from the determination feature value acquisition unit 70.
The microorganism identification unit 71 acquires, from the clustering parameter storage unit 85, the clustering parameters set as a result of the classification in the clustering unit.
The microorganism identification unit 71 further acquires all combinations of the cluster ID and the label from the association storage unit 88.
The microorganism identification unit 71 calculates a cluster ID for each feature generated from the input image 49 using the clustering parameters. The microorganism identification unit 71 calculates a label corresponding to the cluster ID, that is, an identification category to which the microorganism belongs, for each feature amount (step S33).
The microorganism identification unit 71 transmits all the calculated identification classifications to the operation status evaluation unit 28.

The operation status evaluation unit 28 receives from the microorganism identification unit 71 all calculated identification categories.
The operation status evaluation unit 28 evaluates the operation status of the wastewater treatment based on the identification result of the microorganism identification unit 71 (Step S35).

  Next, the effects of the above-described sewage treatment operation status evaluation apparatus and the sewage treatment operation status evaluation method will be described.

The sewage treatment operation status evaluation device 60 according to the second embodiment targets a sewage treatment facility 1 that purifies sewage stored in a biological treatment reaction tank (water treatment tank) 3 with microorganisms, and inhabits the water treatment tank. A sewage treatment operation status evaluation device 60 that analyzes an input image 49 obtained by capturing an image of a microorganism to calculate which classification class the microorganism belongs to, and evaluates the operation status of sewage treatment. The learning data including the learning image 43 relating to the microorganism is input to the machine learning device 90 to perform machine learning, and a microbial feature learning unit 65 that generates an identification parameter relating to identification of the microorganism is used. A microorganism identification unit 71 that calculates an identification category to which the microorganism belongs, and an operation status evaluation unit 2 that evaluates the operation status of the sewage treatment based on the identification result in the microorganism identification unit 71 It has a, and.
According to the above-described configuration, the microbial feature learning unit 65 inputs the learning data including the learning image 43 on the microbe to the machine learning device 90 to perform the machine learning. It is possible to automatically generate an identification parameter that is extracted and reflects this. For this reason, when introducing the sewage treatment operation situation evaluation device 60, no advanced knowledge of image processing is particularly required.
In addition, since there is no need for the worker to intentionally extract the characteristic so that the characteristic inherent to the microorganism is remarkably expressed, highly specialized knowledge about the microorganism is not particularly required. In the second embodiment, it is necessary for the worker to associate a label with each set classified by clustering, but a particularly high level of knowledge required for extracting the characteristics of microorganisms is required. Absent.
Therefore, introduction of the sewage treatment operation status evaluation device 60 is easy.
In addition, the accuracy of the identification parameter can be easily improved by preparing a large amount of learning data to be input to the machine learning device 90. For example, in the second embodiment, the accuracy of the learning parameter of the machine learning device 90 is improved by preparing a large amount of learning data, and the accuracy of the feature amount of the learning image 43 output by the learned model 95 is improved. It affects the accuracy of the clustering parameters, ie the identification parameters. Therefore, it is possible to increase the calculation accuracy of the identification section to which the microorganism belongs, that is, the accuracy of evaluating the operation state of the sewage treatment.

The machine learning device 90 performs convolution deep learning.
According to the above configuration, the sewage treatment operation status evaluation device 60 can be appropriately realized.

The machine learning device 90 executes the machine learning by inputting the learning data and reflecting the characteristics of the microorganism on the learning parameter to generate the learned model 95.
Further, a feature amount extraction unit 68 is provided which inputs the learning image 43 and the input image 49 to the trained model 95 and extracts a feature of the microorganism in the learning image 43 and the input image 49 as a feature amount.
The microbial feature learning unit 65 includes a clustering unit 67 that classifies a plurality of feature amounts corresponding to each of the plurality of learning images 43 into a plurality of sets by clustering, and each of the plurality of sets corresponds to an identification section. And the identification parameters are clustering parameters set as a result of the classification in the clustering unit 67.
In addition, the microorganism identification unit 71 calculates an identification category to which the microorganism belongs by calculating which of the plurality of sets the feature amount of the microorganism in the input image 43 is based on the clustering parameter.
According to the above configuration, the input image 49 used in the evaluation of the driving situation is processed by the learned model 95 in which the characteristics of the microorganisms are reflected, so that the characteristics are efficiently extracted. The effect of the environment can be reduced. Therefore, the operation of the sewage treatment operation status evaluation device 20 is facilitated.
In addition, since the feature amounts of the learning image 43 are accumulated and classified into a plurality of sets by clustering, the classification is calculated by calculating which set the feature amount of the input image 49 corresponds to. High identification becomes possible.

The machine learning device 90 is realized by a convolutional auto encoder.
According to the above-described configuration, when an image is input to the learned model 95 of the machine learning device 90, the image is output as an appropriately compressed feature amount. Therefore, the feature amount is integrated and a clustering process is performed. The amount of data handled at the time is reduced. Therefore, the clustering process becomes easy.

In addition, the operation status evaluation unit 28 evaluates the operation status of the sewage treatment based on the appearance rates of the microorganisms for each of the identification categories calculated by the microorganism identification unit 71 for the plurality of input images 49.
According to the above configuration, the operating state of the sewage treatment can be evaluated with high accuracy.

  In the sewage treatment operation status evaluation device 60, similarly to the first embodiment, even when the types of microorganisms increase, the erroneous evaluation of the operation status is performed because the machine learning device 90 learns to accurately capture the characteristics of the microorganisms. Can be suppressed, and if the learning data is created, the feature extraction itself is automatically performed, so that the increase in the introduction cost is minute. As described above, each of the above-described effects of the sewage treatment operation status evaluation device 60 can be more remarkable when there are many types of microorganisms as compared with the conventional system.

  Note that the sewage treatment operation status evaluation device and the sewage treatment operation status evaluation method of the present invention are not limited to the above-described embodiments and modified examples described with reference to the drawings. Various modifications are possible.

  For example, it goes without saying that the configurations of the machine learning devices 40, 50, and 90 are not limited to those described in the above-described embodiments and modified examples. For example, in the machine learning device 40, more convolutional layers are provided, pooling processing is performed only in a limited number of convolutional layers, one or more intermediate layers are provided between the input layer 47 and the output layer 48, and the like. However, any configuration may be provided as long as it is possible to set such that the characteristics of the microorganisms in the input image can be extracted. The same applies to the machine learning devices 50 and 90.

  Although the control devices 22 and 62 have been described as being realized as one information processing device, each component is divided and mounted on a plurality of information processing devices, and these components operate in cooperation with each other. Needless to say, this may be done. For example, in the control devices 22 and 62, a part for performing learning and a part for evaluating the driving situation may be divided into two corresponding information processing devices.

In addition, the operation status evaluation unit 28 has been described to evaluate whether the operation status is good or deteriorated. However, based on the rate of appearance of microorganisms, the operation status evaluation unit 28 may further evaluate the operation status in, for example, five stages. May be realized.
Alternatively, the evaluation by the driving state evaluation unit 28 may be based on an evaluation axis indicating that the driving state is good or bad, and may be classified into a plurality of states that do not have a priority relationship with each other.

  In addition, the configuration described in each of the above-described embodiments and modified examples can be selected or changed to another configuration as appropriate without departing from the gist of the present invention.

1 wastewater treatment facility 3 biological treatment reaction tank (water treatment tank)
20, 60 Sewage treatment operation status evaluation device 21 Imaging device 22, 62 Control device 23 Microorganism imaging unit 24 Image storage unit 25, 65 Microorganism feature learning unit 26 Determination image acquisition unit 27, 71 Microorganism identification unit 28 Operation status evaluation unit 30 Learning Image acquisition unit 31 Labeling unit 32, 81 Learning processing unit 33, 82 Learning parameter storage unit 40, 50, 50-1 to 50-N, 90 Machine learning device 40A, 95 Trained model 43 Learning image 49 Input image 66 Machine learning Unit 67 clustering unit 68 feature amount extraction unit 69 feature amount accumulation unit 70 decision feature amount acquisition unit 83 learning feature amount acquisition unit 84 clustering execution unit 85 clustering parameter storage unit 86 result presentation unit 88 association storage unit 94 reproduced image

Claims (8)

For a sewage treatment facility that purifies sewage stored in a water treatment tank with microorganisms, an input image obtained by imaging the microorganisms living in the water treatment tank is analyzed to determine to which classification category the microorganisms belong. A sewage treatment operation status evaluation device that calculates and evaluates the operation status of sewage treatment,
A microbial feature learning unit that includes a machine learning device, performs learning by inputting learning data including a learning image regarding the microorganism to the machine learning device, and generates an identification parameter regarding identification of the microorganism.
Using the identification parameters, a microorganism identification unit that calculates the identification classification to which the microorganisms in the input image belong,
An operation status evaluation unit that evaluates the operation status of the sewage treatment based on the identification result in the microorganism identification unit,
A sewage treatment operation status evaluation device comprising:   The sewage treatment operation status evaluation device according to claim 1, wherein the machine learning device performs convolution deep learning. The machine learning device generates, in addition to the learning image, information on the identification classification of the microorganism corresponding to the learning image as teacher data and reflects a feature of the microorganism on a learning parameter to generate a trained model. To perform machine learning,
The sewage treatment operation status evaluation device according to claim 1, wherein the identification parameter is the learning parameter.   The sewage treatment operation status according to claim 1, wherein the machine learning device executes the machine learning by inputting the learning data and reflecting a feature of the microorganism on a learning parameter to generate a learned model. Evaluation device.   A feature amount extraction unit that inputs the learning image and the input image to the trained model and extracts a feature of the microorganism in the learning image and the input image as a feature amount. The sewage treatment operation status evaluation device according to 4.   The sewage treatment operation status evaluation device according to claim 4 or 5, wherein the machine learning device is realized by a convolutional auto encoder.   The said operation condition evaluation part evaluates the operation condition of the said sewage treatment based on the appearance rate of the said microorganisms for every said classification division calculated by the said microorganism identification part with respect to several said input images. The sewage treatment operation status evaluation device according to any one of claims 1 to 6. For a sewage treatment facility that purifies sewage stored in a water treatment tank with microorganisms, an input image obtained by imaging the microorganisms living in the water treatment tank is analyzed to determine to which classification category the microorganisms belong. A sewage treatment operation status evaluation method for calculating and evaluating the operation status of sewage treatment,
Learning data including a learning image of the microorganism is input to a machine learning device to perform machine learning, and an identification parameter related to the identification of the microorganism is generated.
Using the identification parameters, calculate the identification classification to which the microorganisms in the input image belong,
A sewage treatment operation state evaluation method for evaluating an operation state of the sewage treatment based on the identification result.

Images