JP2021103146A - Information processing device, information processing method, control device, water treatment system, and control program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate feature information in which the properties of each floc are accurately reflected. An information processing apparatus (1) captures a series of time-series photographed images of flocs formed by adding a coagulant to a liquid containing suspended solids and stirring the flock, and each image appears in the photographed image. A contour specifying unit (101) for specifying the contour of the flock and a feature information generating unit (103) for generating the feature information of the flock based on the contour are provided. [Selection diagram] Fig. 1

Description

The present invention relates to an information processing device or the like that generates flock feature information that appears in an image.

Conventional technology is to obtain treated water by aggregating suspended solids in a liquid to form flocs and performing solid-liquid separation by adding a chemical such as a flocculant to a liquid containing suspended solids and stirring the liquid. It is used from.

In order to obtain treated water in a stable manner, it is necessary to form flocs of an appropriate size. Therefore, in the step of forming the flocs, it is desirable to perform control for maintaining the size of the flocs at an appropriate size at any time while grasping the properties of the flocs being formed.

Examples of documents in which a technique for grasping the properties of flocs is disclosed include the following Patent Document 1. Patent Document 1 discloses a technique for calculating the degree of cohesion and concentration of flocs as feature information from an image feature amount extracted from a captured image of flocs using a machine learning model.

Japanese Unexamined Patent Publication No. 2019-052985

However, the technique of Patent Document 1 cannot detect individual flocks appearing in an image, so that there is a problem that it is difficult to generate feature information that accurately reflects the properties of each flock. One aspect of the present invention is to realize an information processing apparatus or the like that can generate feature information that accurately reflects the properties of each floc.

In order to solve the above-mentioned problems, the information processing apparatus according to one aspect of the present invention takes a series of time-series photographs of flocs formed by adding a flocculant to a liquid containing suspended solids and stirring the mixture. It includes a contour specifying unit that specifies the contour of each flock appearing in the captured image from the image, and a feature information generating unit that generates the feature information of the flock based on the contour specified by the contour specifying unit.

According to the above configuration, since the flock contour is specified from a series of time-series captured images, it is possible to specify the flock contour with higher accuracy than when the flock contour is specified from a single image. Then, by enabling highly accurate identification of the contour, it becomes possible to generate feature information that accurately reflects the properties of each floc.

In the information processing apparatus according to one aspect of the present invention, the contour specifying portion specifies the contour of each flock reflected in the captured image for each of the series of captured images, and is surrounded by the contour specified by the contour specifying portion. A flock detection unit that detects a region whose position has changed in the series of captured images as a flock is provided, and the feature information generation unit is based on the flock detection result by the flock detection unit. Feature information may be generated.

The contour specified by the contour specifying portion may include contours such as dirt adhering to the object that is the background of the flock, but according to the above configuration, the contour whose position has changed is detected as the flock. , Such things can be prevented from being detected as flocs. Then, according to the above configuration, since the feature information is generated based on such an accurate detection result of the flock, it is possible to generate the feature information that accurately reflects the properties of each flock.

In the information processing apparatus according to one aspect of the present invention, the contour specifying unit is a series of trained models constructed by machine learning the relationship between the image in which the flock is captured and the contour of the flock in the image. For each of the captured images, the outline of the flock appearing in the captured image may be specified.

Since the flock is irregular and multiple flocks may be photographed in an overlapping manner, it is difficult to identify the contour of the flock by normal image processing. However, by using the trained model, the flock can be captured. It becomes possible to specify the contour with high accuracy.

The information processing apparatus according to one aspect of the present invention may include a flock state determination unit that determines a flock formation state using the feature information generated by the feature information generation unit. Since the characteristics of each flock can be accurately reflected in the feature information generated by the feature information generation unit, it is possible to accurately determine the formation state by accurately considering the contour of the flock by using this feature information. ..

The information processing apparatus according to one aspect of the present invention includes a water quality prediction unit that uses the feature information generated by the feature information generation unit to generate water quality information indicating the water quality of the liquid after removing the flocs. You may be. As described above, the characteristic information generated by the feature information generation unit can accurately reflect the properties of each floc. Therefore, by using this feature information, highly accurate water quality prediction that accurately considers the contour of the flocs. Becomes possible.

In the information processing apparatus according to one aspect of the present invention, the feature information generation unit includes (1) the feature information indicating the movement mode of the flock in the series of captured images, and (2) each flock included in the captured image. It may generate at least one of the feature information indicating the uniformity of the size of the above and (3) the feature information indicating the shape of each floc included in the captured image.

If the contour of the flock can be specified with high accuracy, it is possible to generate highly reliable information on each of the above-mentioned feature information. Since these characteristic information is highly related to the properties of flocs, the properties of flocs are accurately grasped by using these characteristic information, and liquids containing suspended solids are appropriately treated. It also becomes possible.

In order to solve the above-mentioned problems, the information processing method according to one aspect of the present invention is an information processing method using one or more information processing devices, in which a flocculant is added to a liquid containing suspended solids and stirred. From a series of time-series captured images of the flock formed by this, the contour specifying step for specifying the contour of each flock reflected in the captured image and the contour of the flock based on the contour specified in the contour specifying step. Includes a feature information generation step to generate feature information. According to this information processing method, the same operation and effect as that of the information processing apparatus can be obtained.

In order to solve the above-mentioned problems, the control device according to one aspect of the present invention forms flocs by adding a flocculant to a liquid containing suspended solids and stirring the liquid, and then solid-liquid separates the liquid. A control device that controls at least one of stirring of the liquid and addition of the coagulant in a water treatment system for obtaining treated water. (1) The liquid is based on feature information generated by the information processing device. It is provided with at least one of a stirring control unit that controls the stirring of the above, and (2) a medication control unit that controls the addition of the flocculant based on the feature information generated by the information processing apparatus. As a result, the feature information of the flock generated by the information processing device can be reflected in the control that affects the formation state of the flock.

In order to solve the above problems, the water treatment system according to one aspect of the present invention forms flocs by adding a flocculant to a liquid containing suspended solids and stirring the liquid, and then separates the liquid into solid and liquid. A water treatment system for obtaining treated water, which includes the information processing device and the control device. This makes it possible to stably obtain treated water while maintaining a state in which flocs are properly formed.

The information processing device and the control device according to each aspect of the present invention may be realized by a computer. In this case, the information is described by operating the computer as each part (software element) included in the information processing device and the control device. A control program of an information processing device that realizes a processing device and the control device by a computer, a control program of the control device, and a computer-readable recording medium on which the control device is recorded are also included in the scope of the present invention.

According to one aspect of the present invention, it is possible to generate feature information that accurately reflects the properties of each floc.

It is a block diagram which shows an example of the main part structure of the information processing apparatus and control apparatus which concerns on Embodiment 1 of this invention. It is a figure which shows the configuration example of the water treatment system which includes the information processing apparatus and the control apparatus. It is a figure which shows the specific example of the contour of the flock from the image which photographed the flock. It is a figure which shows an example of the water quality prediction information for predicting the water quality from the characteristic information. It is a flowchart which shows an example of the process executed by the said information processing apparatus.

[System overview]
The outline of the water treatment system 100 according to the embodiment of the present invention will be described with reference to FIG. FIG. 2 is a diagram showing a configuration example of the water treatment system 100. The water treatment system 100 is a system for solid-liquid separation of a liquid containing suspended solids (for example, water) to obtain treated water.

FIG. 2 shows an information processing device 1, a control device 3, a stirring tank 4, a stirring device 5, a chemical supply device 6, and an imaging unit 7 among the components of the water treatment system 100. In addition to these components, the water treatment system 100 may include components such as a settling tank that setstles and separates solids in a liquid. The stirring tank 4 and the stirring device 5 may be configured as an integrated device (floculator).

The stirring tank 4 is a container that houses the liquid to be processed. A stirring device 5 is provided in the stirring tank 4. The stirring device 5 includes a stirring blade and a motor (not shown), and the liquid in the stirring tank 4 is stirred by driving the stirring blade with the motor. Further, the chemical supply device 6 supplies the chemicals for aggregating the suspended solids dispersed in the liquid to the stirring tank 4. This chemical contains at least a flocculant. As a result, the suspended solids dispersed in the liquid are aggregated in the stirring tank 4 to form the floc F.

The photographing unit 7 is for photographing the flock F in the stirring tank 4. The photographing unit 7 has a configuration in which a support member 73, a background plate 74, a lighting device 75, and a photographing device 76 are arranged on a tubular member 71 having an opening 72 at one end. The support member 73 is a member that supports the background plate 74 inside the tubular member 71, and the background plate 74 is a plate-shaped member that serves as a background when photographing the flock F.

As shown in the drawing, the photographing unit 7 agitates the opening 72 side of the tubular member 71 so that the background plate 74 is parallel to the water surface and the background plate 74 is located at a predetermined depth from the water surface. It is used in a state of being submerged in the tank 4. In this state, the flock F on the background plate 74 is photographed by the photographing device 76 while projecting light by the lighting device 75. Since the flock F is formed in the stirring tank 4, the flock F appears in the photographed image photographed by the photographing apparatus 76.

The method of photographing the flock F is not limited to this example, and for example, a method of submerging a photographing device having a waterproof function in the stirring tank 4 for photographing can be applied, but when the photographing unit 7 is used, the photographing unit 7 is used. It is preferable because it has the following advantages.
(1) Since the tubular member 71 blocks external light, it is possible to take a picture without being affected by external light.
(2) Since the tubular member 71 blocks the wave motion of the water surface, it is possible to take a picture without being affected by the wave motion of the water surface.
(3) The lighting device 75 and the photographing device 76 are not contaminated with liquid, and it is not necessary to provide these devices with a waterproof function.
(4) Since only the flocs above the background plate 74 are photographed, the overlap of the flocs in the depth direction is reduced. This makes it easier to detect individual flocs.
(5) By setting the color of the background plate 74 to a color (for example, white) that makes the contrast with the flock color clear, it is possible to facilitate the detection of flock.
(6) Since the light is projected from a direction inclined with respect to the water surface, the influence of the reflected light on the water surface on the photographing can be reduced.

The information processing device 1 is a device that generates feature information of the flock F based on a series of time-series captured images taken by the photographing device 76. The feature information and its generation method will be described in detail later.

The control device 3 is a device that controls the stirring of the liquid and the addition of the drug. More specifically, the control device 3 controls the operation of the stirring device 5 based on the feature information generated by the information processing device 1 to cause stirring in an manner according to the state of the flock F. Further, the control device 3 controls the operation of the chemical supply device 6 based on the above-mentioned characteristic information, and supplies the chemicals to the stirring tank 4 in an manner according to the state of the flock F.

The control device 3 may be a device that controls only either the stirring of the liquid or the addition of the chemical. Further, instead of providing one control device 3, a control device for controlling the agitation of the liquid and a control device for controlling only the addition of the drug may be provided.

As described above, in the water treatment system 100, a floc F is formed by adding a coagulant to a liquid containing suspended solids and stirring the liquid in the stirring tank 4, and then the liquid is solid-liquid separated and treated. It is a system to obtain water. Further, the flock F formed in the stirring tank 4 is photographed by the photographing unit 7, and the information processing device 1 generates the feature information of the flock F based on a series of time-series photographed images of the flock F. .. Then, the control device 3 controls at least one of stirring of the liquid and addition of the coagulant based on this characteristic information.

As described above, in the water treatment system 100, the characteristic information of the floc F is reflected in the control of stirring and dosing, which affects the formation state of the floc F. Therefore, it is possible to stably obtain water from the liquid to be treated while maintaining the state in which the floc F is properly formed.

[Configuration of information processing device]
A more detailed configuration of the information processing device 1 will be described with reference to FIG. FIG. 1 is a block diagram showing an example of a main configuration of the information processing device 1 and the control device 3. The configuration of the control device 3 will be described later in "Configuration of control device".

As shown in FIG. 1, the information processing device 1 includes a control unit 10 that controls and controls each unit of the information processing device 1, and a storage unit 11 that stores various data used by the information processing device 1. .. Further, the information processing device 1 includes an input unit 12 for receiving an input to the information processing device 1, an output unit 13 for the information processing device 1 to output information, and another device (for example, a control device 3) for the information processing device 1. ) Is provided with a communication unit 14 for communicating with.

Further, the control unit 10 includes a contour specifying unit 101, a flock detection unit 102, a feature information generation unit 103, a flock state determination unit 104, and a water quality prediction unit 105. Then, the image 111 and the contour prediction model 112 are stored in the storage unit 11.

The contour specifying unit 101 identifies the contour of each floc reflected in the captured image from a series of time-series captured images of the flocs formed by adding a coagulant to the liquid containing the suspended solid matter and stirring the flocs. .. The liquid is, for example, industrial wastewater or domestic wastewater. The contour prediction model 112 is used to specify the contour.

The flock detection unit 102 detects a region surrounded by the contour specified by the contour specifying unit 101 as a flock. However, the flock detection unit 102 does not need to detect the entire area surrounded by the contour specified by the contour specifying unit 101 as flock. This is because the contour specified by the contour specifying portion 101 may include contours such as dirt adhering to the background plate 74.

For example, the flock detection unit 102 may detect a region surrounded by the contour specified by the contour specifying unit 101 whose position has changed in a series of captured images as a flock. As a result, even when the contour specified by the contour specifying portion 101 includes a contour such as dirt adhering to the background plate 74, such a contour can be prevented from being detected as a floc.

The feature information generation unit 103 generates flock feature information based on the contour specified by the contour specifying unit 101. In the present embodiment, an example will be described in which the feature information generation unit 103 generates feature information based on the contour specified by the contour specifying unit 101 by using the detection result of the flock detection unit 102. Note that the feature information generation unit 103 may generate feature information using the contour identification result by the contour identification unit 101 instead of the detection result of the flock detection unit 102.

The flock state determination unit 104 determines the flock formation state using the feature information generated by the feature information generation unit 103. The above-mentioned feature information is generated based on a specific result of contours from a series of time-series captured images of flocs, and the properties of each floc can be accurately reflected in this feature information. Therefore, by using this feature information, it is possible to accurately determine the formation state in consideration of the contour of the flock.

The water quality prediction unit 105 uses the feature information generated by the feature information generation unit 103 to generate water quality information indicating the water quality of the treated water, which is a liquid after removing flocs. As described above, the above-mentioned feature information is generated based on the specific result of the contour from a series of time-series captured images of the flock, and the characteristic information of each flock accurately reflects the properties of each flock. Can be done. Therefore, by using this feature information, it is possible to predict the water quality with high accuracy in consideration of the contour of the floc.

Image 111 is an image obtained by photographing the flock. The image 111 may be a moving image or a still image. In the former case, the contour specifying unit 101 acquires a frame image from the image 111 and sets it as a series of time-series captured images. In the latter case, a series of time-series captured images of the flock may be stored as the image 111.

The contour prediction model 112 is a trained model constructed by machine learning the relationship between the image in which the flock is captured and the contour of the flock in the image. As described above, the contour specifying unit 101 uses the contour prediction model 112 to specify the contour of the flock appearing in the captured image for each of the series of captured images.

Since the flock is irregular and multiple flocks may be photographed in an overlapping manner, it is difficult to identify the contour of the flock by normal image processing, but by using the contour prediction model 112, the flock can be captured. It becomes possible to specify the contour with high accuracy.

As described above, the information processing apparatus 1 has a contour specifying unit 101 that specifies the contour of each flock reflected in the captured image from a series of time-series captured images of the flock, and a contour that is specified by the contour specifying unit 101. It is provided with a feature information generation unit 103 that generates feature information of flock based on the feature information. According to this configuration, since the flock contour is specified from a series of time-series captured images, it is possible to specify the flock contour with higher accuracy than when the flock contour is specified from a single image. Then, by enabling highly accurate identification of the contour, it is also possible to generate feature information that accurately reflects the properties of each floc.

[Control device configuration]
A more detailed configuration of the control device 3 will be described with reference to FIG. As shown in FIG. 1, the control device 3 includes a control unit 30 that controls each unit of the control device 3 in an integrated manner, and a storage unit 31 that stores various data used by the control device 3. Further, the control device 3 communicates with an input unit 32 that receives an input to the control device 3, an output unit 33 for the control device 3 to output information, and the control device 3 communicates with another device (for example, an information processing device 1). It is provided with a communication unit 34 for the purpose of processing. Further, the control unit 30 includes a feature information acquisition unit 301, a stirring control unit 302, and a medication control unit 303. Either the stirring control unit 302 or the medication control unit 303 may be omitted.

The feature information acquisition unit 301 acquires the feature information generated by the information processing device 1. The feature information may be acquired, for example, by communication via the communication unit 34, or may be acquired by the user of the control device 3 via the input unit 32.

The agitation control unit 302 controls the agitation of the liquid based on the characteristic information acquired by the characteristic information acquisition unit 301. More specifically, the stirring control unit 302 controls the operation of the stirring device 5 based on the above-mentioned characteristic information, so that the stirring is performed in a mode according to the flock state. The control content according to the feature information may be determined in advance. For example, the stirring control unit 302 may reduce the stirring speed of the stirring device 5 when the above characteristic information indicates that the size of the flocs is too small.

The dosing control unit 303 controls the addition of the flocculant based on the feature information acquired by the feature information acquisition unit 301. More specifically, the dosing control unit 303 controls the operation of the drug supply device 6 based on the above-mentioned feature information, so that the drug is supplied to the stirring tank 4 in an manner according to the flock state. The control content according to the feature information may be determined in advance. For example, the dosing control unit 303 may reduce the dosage when the above characteristic information indicates that the size of the flocs is excessive or the viscosity of the flocs is high. ..

In the case of a water treatment system in which the operator manually adjusts the dosage and the like, the medication control unit 303 notifies the output unit 33 of the content of the adjustment to be performed by the operator, thereby performing the work. Dosing may be controlled through a person. The same applies to the control of stirring.

As described above, the control device 3 includes a stirring control unit 302 that controls the stirring of the liquid based on the feature information generated by the information processing device 1, and a medication control unit 302 that controls the addition of the flocculant based on the feature information. It is equipped with 303. Therefore, the characteristic information of the flocs can be reflected in the control that affects the formation state of the flocs, whereby the state in which the flocs are properly formed can be maintained, and water can be stably discharged from the liquid to be treated. It will be possible to obtain.

[Example of contour identification and flock detection]
An example of contour identification and flock detection will be described with reference to FIG. FIG. 3 is a diagram showing a specific example of the contour of the flock from the captured image of the flock. In FIG. 3, the contour specified by the contour specifying portion 101 is shown by a broken line.

More specifically, FIG. 3 shows images 111a, 111b, and 111c, the imaging times of which are t1, t2, and t3, respectively. Note that t1 <t2 <t3, and the time from time t1 to t2 is the same as the time from time t2 to t3. This time may be set to such a time that the corresponding flock can be specified between the images, and may be set to, for example, about 0.1 second.

When detecting flock from the image 111a, the flock detection unit 102 first detects the regions a1 to a9 surrounded by the contour line as flock candidates. Next, the flock detection unit 102 determines whether or not the regions a1 to a9 include those whose positions do not change, and if they are included, excludes them from the candidates.

For example, the flock detection unit 102 may exclude images 111a and 111b having the same position and an area difference within a predetermined range from the candidates. In the example of FIG. 3, the region a1 of the image 111a is excluded because it has the same position and size as the region b1 of the image 111b. On the other hand, the flock detection unit 102 does not exclude candidates that are not detected in the image 111b such as the area a5 and candidates whose positions have changed, and detects those areas as flock.

In addition, when processing a liquid, a plurality of flocs may gather to form an agglomerated portion of the flocs. For example, the image 111a of FIG. 3 includes an agglomerated portion A1 composed of three regions a2 to a4 and an agglomerated portion A2 composed of four regions a6 to a9. The detection of the contour of the flock contained in the agglomerated portion is more difficult than the detection of the contour of the flock other than the agglomerated portion, and the detection omission of the contour is likely to occur. Therefore, the flocs contained in the agglomerated portion may be detected by the following treatment.

First, the flock detection unit 102 detects the aggregated portion from the image 111a. For example, the flock detection unit 102 may specify a group of a plurality of adjacent regions as an agglomeration unit. Subsequently, the flock detection unit 102 confirms the validity of detecting each region included in the agglomerated portion as a flock.

In confirming the validity, first, the flock detection unit 102 detects the agglomerate portion B1 corresponding to the agglomerate portion A1 detected in the image 111a from the image 111b which is a subsequent image of the image 111a in chronological order. The corresponding agglomerated portion B1 may be detected based on the position, size, or the like. Similarly, the flock detection unit 102 detects the agglomerated portion B2 corresponding to the agglomerated portion A2.

Then, the flock detection unit 102 compares the number of regions contained in the agglomerated portions A1 and B1 and also compares the number of regions contained in the agglomerated portions A2 and B2. Here, if the number of regions is the same or the number of regions included in the aggregated portion on the image 111a side is large, the result of specifying the contour in the aggregated portion on the image 111a side is considered to be valid. 102 may detect each region included in the agglomerated portion on the image 111a side as a floc.

For example, the agglomerated portion A2 includes four regions a6 to a9, while the agglomerated portion B2 corresponding to the agglomerated portion A2 contains three regions b7 to b9. Therefore, the flock detection unit 102 detects the regions a6 to a9 included in the agglomerated portion A2 as flock for the agglomerated portion A2.

On the other hand, the number of regions included in the agglomerated portion A1 is smaller than that of the agglomerated portion B1, and there is a possibility that contour detection omission has occurred in the agglomerated portion A1. Therefore, the flock detection unit 102 may increase the number of regions by dividing the region included in the aggregated portion A1 with respect to the aggregated portion A1 having a smaller number of regions.

Which region is to be divided may be determined based on the positional relationship, shape, size, etc. of the region. For example, the flock detection unit 102 divides a region a3 having a shape in which two circles are connected into an upper portion and a lower portion from the connection portion of the two circles in the region included in the agglutinating portion A1 in FIG. , Regions b4 and b5, respectively, may be set and these may be detected as flocs.

By the above processing, the flock detection unit 102 can detect the upper parts of a2 and a3, the lower part of a3, and a4 to a9 as flock in the regions a1 to a9 detected from the image 111a. Further, the flock detection unit 102 can detect flock from the image 111b by performing the same processing on the image 111b and the image 111c.

Flock may be detected using three or more of a series of time-series images. As a result, the flock detection accuracy can be further improved. For example, in the example of FIG. 3, the number of regions included in the agglomerated portion A1 among the agglomerated portions A1, B1 and C1 corresponding to each other is different from that of the other two agglomerated portions. Specifically, the agglomerated portions B1 and C1 each include four regions, whereas the agglomerated portion A1 contains three regions. In this case, since it is considered that the result of specifying the contour in the agglomerated portion A1 is incorrect, the flock detection unit 102 sets the region included in the agglomerated portion A1 so that a region similar to the agglomerated portions B1 and C1 is formed. It may be divided.

Further, the flock detection unit 102 may exclude the region included in the agglomeration part from the flock detection target. In this case, the flock detection unit 102 detects an independent region that is not adjacent to the other region as a flock. For example, in the image 111a of FIG. 3, there are two independent regions a1 and a5 that are not adjacent to the other regions, and the position of the region a1 has not changed. Therefore, the flock detection unit 102 can use the region 102. Only a5 is detected as a floc. Since the flock independent of the other flock moves without being affected by the other flock, this configuration makes it possible to more accurately identify the movement speed and movement pattern of the flock.

[About contour prediction model]
The contour prediction model 112 may be constructed by machine learning so that the contour as shown by the broken line in FIG. 3 can be specified. For this machine learning, teacher data in which information indicating the outline of the flock in the image is associated with the image in which the flock is captured as correct answer data may be used. It is preferable that the contour to be the correct answer data is set by a person visually observing the above image, at least in part. This makes it possible to detect contours that are close to human senses.

As the learning algorithm, any algorithm can be applied as long as it can construct a trained model that can output information showing the outline of the flock by inputting an image. For example, the contour prediction model 112 may be a trained model such as a convolutional neural network.

[Example of feature information to be generated]
The feature information generation unit 103 includes (1) feature information indicating the movement mode of the flock in a series of captured images, (2) feature information indicating the uniformity of the size of each flock included in the captured image, and (3) the captured image. At least one of the feature information indicating the shape of each floc contained in may be generated.

If the contour of the flock can be specified with high accuracy, it is possible to generate highly reliable information on each of the above-mentioned feature information. Since these characteristic information is highly related to the properties of the flocs, it is possible to accurately grasp the properties of the flocs by using these characteristic information and appropriately process the liquid. ..

The feature information in (1) above may indicate, for example, the average moving speed of flocs. In this case, the feature information generation unit 103 may calculate the average moving speed from the moving amount and moving time of each flock detected by the flock detecting unit 102 from each of the series of captured images, and use this as the feature information. The amount of movement may be calculated from the change in the position of the flock between consecutive captured images. Further, the difference in shooting time of consecutive shot images may be set as the moving time.

Further, the feature information in (1) above may indicate, for example, a flock movement pattern. As the movement pattern, for example, a type such as "straight ahead" or "random" may be set in advance, and information indicating which of these types corresponds may be used as feature information. Further, which type is applicable may be specified based on the moving direction of the flock between the captured images. For example, the position of the flock in a series of captured images (three or more) can be represented by the coordinates in the captured image plane, and a vector indicating the moving direction of the flock between the captured images can be calculated.

The feature information generation unit 103 calculates such a vector for each floc, and when the flocs of the vector having the same or similar directions exist in a predetermined ratio or more with respect to all the flocs, it corresponds to the type of “straight ahead”. The feature information indicating the above may be generated. On the other hand, the feature information generation unit 103 may generate feature information indicating that it corresponds to the "random" type when it does not correspond to this type.

The feature information in (1) above may be any information indicating the movement mode of the flock, in other words, the time-series change of the flock, and is not limited to each of the above examples. For example, the feature information generation unit 103 may use the feature information indicating the vibration, rotation, or movement locus of the flock as the feature information of the above (1).

The feature information in (2) above may be, for example, the standard deviation of the area of the flocs. Further, for example, the uniformity of the size of the flocs may be categorized in advance such as "uniform" and "non-uniform", and the information indicating which type corresponds to the characteristic information may be used as the feature information. Which type is applicable may be determined based on the standard deviation of the area of the flocs and the like.

Regarding the feature information of (3) above, a type of flock shape such as "circular" or "indeterminate form" may be set in advance, and information indicating which of them corresponds may be used as the feature information. .. For example, the feature information generation unit 103 may determine whether or not each flock corresponds to a "circle" by analyzing the shape of the contour of each flock. Then, the feature information generation unit 103 may generate feature information indicating that the flock corresponding to the “circular” corresponds to the “circular” type when the flock corresponding to the “circular” is present in a predetermined ratio or more with respect to all the flocks. On the other hand, the feature information generation unit 103 may generate feature information indicating that it corresponds to the "indeterminate form" type when it does not correspond to this type.

The feature information generated by the feature information generation unit 103 may be any information related to the properties of the flocs or the water quality of the treated water obtained from the flocs, and is not limited to each of the above examples. For example, the feature information generation unit 103 may generate feature information indicating the size of the flock (for example, the average area of the flock), or may generate the distribution mode of the flock (for example, uniformly distributed / non-uniformly distributed) in the captured image. The information shown may be generated.

Further, for example, the feature information generation unit 103 may generate feature information indicating a growth mode of flocs. In this case, the feature information generation unit 103 generates feature information based on the detection result of flock from each of the plurality of captured images taken at a predetermined time interval. For example, the feature information generation unit 103 may use the rate of increase in the average area of the flocs as the feature information. The predetermined time is such that the growth of flocs can be observed.

The above-mentioned feature information is used for predicting the water quality by the water quality prediction unit 105 and determining the flock formation state in the flock state determination unit 104, and is also transmitted to the control device 3 via the communication unit 14 to stir. And used to control medication.

[Example of water quality prediction method]
An example of a water quality prediction method by the water quality prediction unit 105 will be described with reference to FIG. FIG. 4 is a diagram showing an example of water quality prediction information for predicting water quality from characteristic information. The water quality prediction information shown in FIG. 4 is table-type information in which a plurality of types of characteristic information and water quality information indicating the water quality of the treated water after solid-liquid separation are associated with each other. Specifically, one water quality information is associated with a combination of four types of characteristic information of "movement speed", "movement pattern", "size", and "distribution".

"Movement speed" is characteristic information indicating how fast each flock is moving. In the example of FIG. 4, the "moving speed" is represented by two stages of "fast" and "slow". "Fast" and "slow" may be classified based on, for example, the average moving speed of flock.

The "movement pattern" is characteristic information indicating how the flock is moving. As described above, "straight ahead" and "random" can be classified based on a vector indicating the moving direction of each flock.

"Size" is characteristic information indicating the size of the flock. In the example of FIG. 4, "size" is represented by two stages of "large" and "small". "Large" and "small" may be classified based on, for example, the average area of flocs.

"Distribution" is characteristic information indicating a distribution mode of flock in a photographed image. In the example of FIG. 4, the "distribution" is represented by two stages of "uniform" and "non-uniform". “Uniform” and “non-uniform” may be classified based on, for example, the standard deviation of the flock density in each region when the captured image is divided into a plurality of regions.

The water quality information indicates the water quality of the treated water after removing the flocs by solid-liquid separation from the liquid on which the flocs having the properties represented by the combination of the above characteristic information are formed. In the example of FIG. 4, the water quality information is either "good" or "bad". "Good" and "bad" may be classified based on, for example, the turbidity of the treated water obtained from the liquid in which the flocs having the properties represented by the combination of each characteristic information are formed, or the treatment. The state of water may be classified based on the judgment result judged by observing the state of water. In addition, if the water quality of the treated water obtained from the liquid in which the flocs having the properties represented by the combination of each characteristic information is formed is empirically known, the water quality information is set based on such experience. You may leave it.

When the water quality prediction unit 105 predicts the water quality using the water quality prediction information of FIG. 4, the feature information generation unit 103 generates each of the above feature information. Then, the water quality prediction unit 105 specifies the water quality information associated with the generated combination of the characteristic information in the water quality prediction information. The water quality indicated by the specified water quality information is the prediction result of the water quality prediction unit 105.

The method of using the water quality prediction result is not particularly limited. For example, the water quality prediction unit 105 may output the water quality information to the output unit 13 or may transmit the water quality information to the control device 3 via the communication unit 14. .. In the latter case, the stirring control unit 302 of the control device 3 may control the stirring according to the received water quality information. For example, if the received water quality information indicates that the water quality is "bad", the stirring control unit 302 may change the stirring speed. How to change the stirring speed may be determined based on various feature information acquired by the feature information acquisition unit 301. Similarly, if the received water quality information indicates that the water quality is "poor", the dosing control unit 303 may change the dosage, the dosing rate, or the type of drug to be charged.

[Example of method for determining flock formation state]
The flock state determination unit 104 can determine the flock formation state in the same manner as the water quality prediction by the water quality prediction unit 105 described above. That is, the formation state determination information in which a plurality of types of feature information and the flock formation state are associated with each other may be prepared in advance. As a result, the flock state determination unit 104 can determine the flock formation state from the feature information generated by the feature information generation unit 103 by using the formation state determination information.

The flock formation state may be represented by, for example, "good", "bad" or the like. This classification may be based on a characteristic value related to the processability of the floc, such as the average sedimentation speed of the floc having a property represented by a combination of each characteristic information, or the state of the floc. May be classified based on the judgment result that a person observes and judges.

In addition, when the flock formation state is poor, "excessive aggregating agent", "insufficient aggregating agent", "excessive agitation rate", "insufficient agitation rate", etc. are used so that a specific countermeasure can be easily specified. It may be further divided. In this case, each division may be associated with a special quantity characteristic of the flock in the formed state corresponding to the division. For example, if it is empirically known that flocs when the coagulant is excessive have a slow moving speed and the distribution is often uneven, the "moving speed" is opposed to the "coagulant excess". The characteristic information that the speed is slow and the "distribution" is non-uniform may be associated.

The method of using the flock formation state determination result is not particularly limited. For example, the flock state determination unit 104 may output information indicating the flock formation state to the output unit 13 or may output the information indicating the flock formation state to the output unit 13 or via the communication unit 14. It may be transmitted to the control device 3. In the latter case, the stirring control unit 302 of the control device 3 may control the stirring according to the received information. For example, if the received information indicates that the flock formation state is "bad", the stirring control unit 302 may change the stirring speed. How to change the stirring speed may be determined based on various feature information acquired by the feature information acquisition unit 301. Similarly, if the received information indicates that the floc formation state is "bad", the dosing control unit 303 may change the dosage, the dosing rate, or the type of drug to be added.

[Other examples of water quality prediction method and floc formation state determination method]
If a prediction model that models the relationship between the feature information generated by the feature information generation unit 103 and the water quality information is constructed, the water quality prediction unit 105 can derive the water quality information using the prediction model. For example, a predictive model that models the relationship between the average moving speed, average area, and density of flocs and the turbidity of the treated water after solid-liquid separation may be used.

In this case, the water quality prediction unit 105 can calculate the predicted value of the turbidity of the treated water by inputting the average moving speed, the average area, and the density of the flocs generated by the feature information generation unit 103 into the prediction model. This value can be output as water quality information.

Similarly, if a determination model that models the relationship between the feature information generated by the feature information generation unit 103 and the flock formation state is constructed, the flock state determination unit 104 uses the determination model to form the flock. The state can be determined. Both the judgment model and the prediction model as described above can be constructed by machine learning.

[Processing flow]
The flow of the process (information processing method) executed by the information processing apparatus 1 will be described with reference to FIG. FIG. 5 is a flowchart showing an example of processing executed by the information processing device 1. Before the start of the process of FIG. 5, the imaging unit 7 is used to photograph the liquid in which the flocs are formed, and a series of captured time-series images are stored in the storage unit 11 of the information processing device 1 as an image 111. It is assumed that it is remembered as.

In S1 (contour specifying step), the contour specifying unit 101 identifies the contour of each flock reflected in the image 111 from the image 111. Specifically, the contour specifying unit 101 inputs each of a series of time-series images 111 into the contour prediction model 112, and identifies the contour of each flock from the data output by the contour prediction model 112.

In S2, the flock detection unit 102 detects the flock that appears in the image 111 based on the specific result of S1. At this time, the flock detection unit 102 may detect a region surrounded by the contour specified in S1 whose position has changed in a series of images 111 as a flock. Further, the flock detection unit 102 is independent of other regions, and may detect only a region that does not form an agglomerated portion as a flock.

In S3 (feature information generation step), the feature information generation unit 103 generates feature information based on the detection result of S2. As described above, since the detection of S2 is performed based on the contour specified in S1, it can be said that the feature information is generated in S3 based on the contour specified in S1.

In S4, the flock state determination unit 104 determines the flock formation state using the feature information generated in S3. The method for determining the flock formation state is as described above, and the description will not be repeated here.

In S5, the water quality prediction unit 105 uses the feature information generated in S3 to generate water quality information indicating the water quality of the treated water after removing the flocs. The method of generating water quality information is as described above, and the description is not repeated here. The process of S5 may be performed before the process of S4, or these processes may be performed in parallel.

By the above processing, the characteristic information, the information indicating the formation state, and the water quality information of the flocs appearing in the series of time-series images 111 are generated. Although not shown, some or all of this information may be transmitted to the control device 3, which allows the control device 3 to perform agitation and dosing control according to the information. .. The above processing is preferably performed at predetermined time intervals. This makes it possible to stably continue the water treatment in the water treatment system 100.

[Other examples of contour detection]
In the above example, each of the series of captured images is input to the contour prediction model 112 to specify the contour, but the contour is specified by inputting a plurality of captured images to the contour prediction model 112 at once. May be good. In this case, as the contour prediction model 112, a model constructed by associating a plurality of captured images with correct answer data indicating contours to be specified from the captured images may be used. Similarly, it is also possible to input a moving image into the contour prediction model 112 to specify the contour. Even when these configurations are adopted, it is possible to specify the contour with high accuracy in consideration of the time-series change of the flock, as in the above example.

[Other examples of system configuration]
In the above embodiment, the example in which the information processing device 1 and the control device 3 are independent devices has been described, but these may be combined into one device. Further, a part of the processing executed by each of these devices may be executed by another information processing device other than these devices. For example, the cloud server may be made to specify the contour using the contour prediction model 112. In this case, the contour specifying unit 101 identifies the contour of each flock reflected in the image 111 by transmitting the image 111 to the cloud server and receiving the contour specifying result from the cloud server. As described above, the execution subject of each treatment described in each of the above embodiments can be appropriately changed, and the water treatment system 100 according to the present invention can be realized in various system configurations.

[Example of realization by software]
The control blocks (particularly, each unit included in the control unit 10 and the control unit 30) of the information processing device 1 and the control device 3 may be realized by a logic circuit (hardware) formed in an integrated circuit (IC chip) or the like. However, it may be realized by software.

In the latter case, the information processing device 1 and the control device 3 include a computer that executes an instruction of a control program, which is software that realizes each function. The computer includes, for example, one or more processors and a computer-readable recording medium that stores the control program. Then, in the computer, the processor reads the control program from the recording medium and executes it, thereby achieving the object of the present invention. As the processor, for example, a CPU (Central Processing Unit) can be used. As the recording medium, a "non-temporary tangible medium", for example, a ROM (Read Only Memory) or the like, a tape, a disk, a card, a semiconductor memory, a programmable logic circuit, or the like can be used. Further, a RAM (Random Access Memory) or the like for expanding the control program may be further provided. Further, the control program may be supplied to the computer via an arbitrary transmission medium (communication network, broadcast wave, etc.) capable of transmitting the control program. It should be noted that one aspect of the present invention can also be realized in the form of a data signal embedded in a carrier wave, in which the control program is embodied by electronic transmission.

The present invention is not limited to the above-described embodiments, and various modifications can be made within the scope of the claims, and the embodiments obtained by appropriately combining the technical means disclosed in the different embodiments. Is also included in the technical scope of the present invention.

1 Information processing device 101 Contour identification unit 103 Feature information generation unit 104 Flock state determination unit 105 Water quality prediction unit 112 Contour prediction model 3 Control device 100 Water treatment system

Claims (10)

From a series of time-series captured images of flocs formed by adding a flocculant to a liquid containing suspended solids and stirring, a contour specifying part that identifies the contour of each floc reflected in the captured image, and
An information processing device including a feature information generation unit that generates feature information of the flock based on the contour specified by the contour specifying unit. The contour specifying unit identifies the contour of each flock reflected in the captured image for each of the series of captured images.
It is provided with a flock detection unit that detects as a flock a region in which the position of the contour specifying portion is changed in the series of captured images among the regions surrounded by the specified contour.
The information processing device according to claim 1, wherein the feature information generation unit generates the feature information based on the detection result of the flock by the flock detection unit. The contour specifying portion is captured in the captured image for each of the series of captured images by using a learned model constructed by machine learning the relationship between the image in which the flock is captured and the contour of the flock in the image. The information processing apparatus according to claim 1 or 2, which specifies the contour of a flock. The information processing apparatus according to any one of claims 1 to 3, further comprising a flock state determination unit for determining a flock formation state using the feature information generated by the feature information generation unit. Claim 1 to any one of claims 1 to 4, further comprising a water quality prediction unit that generates water quality information indicating the water quality of the liquid after removing the flocs by using the feature information generated by the feature information generation unit. The information processing device described. The feature information generation unit
(1) The feature information indicating the movement mode of the flock in the series of captured images.
(2) The feature information indicating the uniformity of the size of each flock included in the photographed image, and (3) the feature information indicating the shape of each flock included in the photographed image.
The information processing apparatus according to any one of claims 1 to 5, which generates at least one of the above. An information processing method using one or more information processing devices.
From a series of time-series captured images of flocs formed by adding a flocculant to a liquid containing suspended solids and stirring, a contour specifying step for identifying the contour of each floc reflected in the captured image, and
An information processing method including a feature information generation step of generating feature information of the flock based on the contour specified in the contour specifying step. After forming flocs by adding a flocculant to a liquid containing suspended solids and stirring the liquid, stirring the liquid and adding the flocculant in a water treatment system for solid-liquid separation of the liquid to obtain treated water. A control device that controls at least one of
(1) A stirring control unit that controls stirring of the liquid based on the feature information generated by the information processing apparatus according to any one of claims 1 to 5.
(2) A control device including at least one of a medication control unit that controls the addition of the flocculant based on the feature information generated by the information processing device according to any one of claims 1 to 5. .. A water treatment system in which a floc is formed by adding a flocculant to a liquid containing suspended solids and stirring the liquid, and then the liquid is solid-liquid separated to obtain treated water.
The information processing device according to any one of claims 1 to 6.
A water treatment system including a control device that controls at least one of stirring of the liquid and addition of the flocculant based on the feature information generated by the information processing device. A control program for operating a computer as the information processing device according to claim 1, wherein the computer functions as the contour specifying unit and the feature information generating unit.

Images