JPH0614006B2 - Flock image recognition device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a flock image recognition device for a water purification plant, and more particularly to a flock image recognition device for automatically determining a threshold for binarizing a flock image.

[Field of Application of the Invention] In a water purification plant, since the particle size of suspended solids in raw water is small, these are aggregated to form aggregates (flocs), and the process of settling the flocs is performed. Therefore, it is essential to monitor flocs in the floc formation pond (mixing pond).

Conventionally, the maintenance of the water purification plant has been visually monitored several times a day to monitor the flocs. Since it depends on visual observation, there is a drawback that the judgment criteria are subjective and qualitative, and the monitoring result is difficult to be reflected in the driving operation. Further, since the monitoring frequency is discontinuous, there is a drawback that countermeasures against agglomeration failure are postponed and the trouble becomes serious.

On the other hand, recently, industrial TV cameras (ITV)
Is used to monitor the flocs in the floc formation pond. However, even in this case, monitoring is still subjective and discontinuous because it depends on human vision.

In order to improve such a defect, as described in Japanese Patent Application Laid-Open No. 54-143296, there has been proposed a method of using a photoelectric conversion device to extract an electric signal according to the shape of a floc. However, if there is a change in water quality (turbidity) or a change in ambient light, it is difficult to obtain a stable flock size. If these problems are not solved, unattended monitoring using only a monitoring device is practical. Cannot be used for.

[Object of the Invention]

It is an object of the present invention to provide a flock image recognition device which enhances the accuracy of flock image recognition by appropriately setting a threshold value for binarization when recognizing a flock group in a flock formation pond (mixing pond). Is.

[Outline of Invention]

According to the present invention, the threshold for binarization is automatically determined according to the number of flocs in the binarized image and the average area of the flocs.

The details will be described below.

Generally, when performing image processing using an image processing device,
A problem is how to set a threshold value for binarization.
Fixed binarization can be considered in which the threshold value is set while binarizing the grayscale image recognized by the underwater camera and comparing the binary image, but in the case of fixed binarization, the ambient light of the underwater camera is set after the threshold value is set. If there is a change in turbidity or fluctuation in turbidity, 2
The value image may not be binarized well as compared with the grayscale image, and unless the threshold setting is changed, the correct image processing calculation is not performed. According to the present invention, the setting of the threshold value for binarization is automatically determined by obtaining the number of objects and the average area of the objects in the binary image so that the correct image processing operation can always be performed. is there.

Example of Invention

Figure 2 shows the block diagram of the floc formation pond of the water purification plant. The flock formation pond 10 has three flock formation ponds 10A and 10A.
The inflow water 10a, which is composed of B and 10C and includes the minute floc group that has flowed in from the rapid mixing tank indicated by the arrow, sequentially flows through the floc formation ponds 10A → 10B → 10C and flows out as the outflow water 10b.

In each of the floc formation ponds 10A to 10C, a stirring paddle 1
1A to 11C, the inflowing floc groups are stirred, and the flocs are agglomerated by colliding with each other. A rectification wall 13A is provided between the flock formation basins 10A and 10B, and a rectification wall 13B is provided between the flock formation basins 10B and 10C to perform rectification. The particle size of the flocs 12 gradually increases each time the flocs forming ponds 10A → 10B → 10C pass.

To monitor the flock, a flock imaging device including a TV camera is installed in any of the formation ponds 10A to 10C of the flock formation pond 10. Which of the flock formation ponds 10A to 10C is to be installed depends on the purpose of floc recognition. For the purpose of monitoring the final degree of floc formation, it is installed in the final floc formation pond 10C (that is, before the sedimentation pond), and for the purpose of monitoring the process of floc formation, it is installed in the floc formation pond 10A or 10B. .

FIG. 3 shows an embodiment of a flock image recognition device including a flock imaging device. In FIG. 3, an airtight container 20, a TV camera 30 provided inside the container 20 and a lighting device 40 are installed in the flock formation pond 10.

An observation window 21 is provided on the front surface of the airtight container 20, and a wiper 22 for preventing flock from attaching to the window 21 is provided in front of the window 21. The wiper 22 is driven by a motor, but its mechanism is omitted in the figure. The back screen 24 is fixed to support members 24A and 24B supported by the airtight container 20.
Has a black color and plays a role of improving the contrast so that the TV camera 30 can image the flock existing in the space between the observation window 21 and the screen 24 with a good visibility. In particular, the color of the flock 12 is white, and the TV screen 3 of the flock 12 is selected by making the back screen 24 black.
The contrast viewed from 0 is extremely high.

A lighting device 40 as an auxiliary means for increasing the contrast,
Further, the light shielding cover 42 is provided. The reason for this is as follows.
The flock formation pond 10 is open to the atmosphere for the purpose of constantly monitoring the flock. Therefore, the amount of light incident on the flock formation pond 10 changes with the passage of time, and the weather is strongly photographed. A lighting device 40 is usually installed to constantly monitor the flocs. Needless to say, the installation of the lighting device 40 is sufficient and the installation of the lighting device 40 may be performed on the flock formation pond 10 for the purpose of mere monitoring relying on the eyes of the maintenance manager.

However, even if the lighting device 40 is installed, the change in ambient illuminance strongly affects the image recognition accuracy of the floc group. For example, if the illuminance is low, the flock is recognized as small, and conversely, if the illuminance is high, the flock is recognized as large. In order to remove this effect, it is necessary not to be influenced by changes in illuminance as a natural phenomenon. In this embodiment, a light-shielding cover 42 is provided to darken the surroundings so that the illuminance of only the illumination device 40 is constant.

The ITV 30 fixed in the airtight container 20 is a close-up lens 3
1. Observation window 21 made of transparent material such as glass
The image of the flock 12 in the flock formation pond 10 is enlarged and recognized through. The wiper 22 operates periodically to remove stains on the surface of the observation window 21.

The baffles 25A, 25B and 25C were installed to prevent light scattering. Furthermore, the T of the back screen 24
A reference white portion for checking the illumination device is provided on the observation surface from the V camera 30. While observing the reference white part with the TV camera, the illumination device 40 is judged to be normal,
If it cannot be observed, it is determined to be abnormal.

In FIG. 3, the image capturing circuit 32, the illumination control circuit 41, the image recognition processing unit 50, and the image recognition control device 60 are provided outside the above-described imaging device. Lighting control device 40
The control circuit 41 instructs the TV camera 3 to illuminate
The TV control circuit 32 takes in the captured image by 0.
The output of this circuit 32 becomes a video signal.

FIG. 1 shows an embodiment of the image recognition processing unit 50 and the image recognition control device 60 according to the present invention.

The image recognition processing unit 50 extracts information useful for water quality management of the water purification plant from the image information obtained by the TV camera 30,
Various treatments such as particle size and distribution of flocs are performed. The controller 60 controls those timings.

In FIG. 1, the image recognition processing unit 50 has the following configuration.

Automatic threshold value determining device 90 ... A device for automatically determining a threshold value, and information for that is received from the memory 108.

Abnormality detection device 91 ... Obtains data for determining whether the illumination device 40 is abnormal (turned off).

Alarm output determination unit 92 ... Determines whether or not an alarm is output based on the data obtained by the abnormality detection device 91.

Abnormality alarm unit 93 ... Provides an alarm when an abnormality is determined.

Wiper drive unit 94: Performs wiper drive. When an abnormality is detected, the wiper driving frequency is increased to re-recognize whether or not there is an abnormality.

Recognition timing control unit 100: Captures a video signal from the control circuit 32.

AD converter 101 ... AD-converts a video signal.

Threshold input circuit 102 ... Takes in and outputs the threshold determined by the automatic threshold determination circuit 90.

Binarization circuit 103 ... The multi-valued output of the AD converter 101 is binarized based on the output threshold of the threshold input circuit 102.

Labeling circuit 104 ... Labels (numbers) the flock within one screen.

Particle size measuring circuit 106 ... Particle size is measured for each floc.

Particle size calculation mode designating circuit 105 ... Designates a particle size calculation mode.

Particle size comparison circuit 107 ... Classify according to particle size for each floc.

Memory 108 ... Having memory elements 108A to 108Z,
The memory element 108 corresponding to each size corresponds to the size of the particle size.
It is distributed to A to 108Z and stored. The stored data is the number.

Recognition number storage unit 109 ... Performs image recognition a plurality of times. At that time, the flock image to be recognized each time is a flock image captured in a time-division manner within a fixed time period.

Particle size distribution calculation display unit 100: Calculates and displays the particle size distribution of flocs.

The recognition timing control circuit 100 shown in FIG.
The time interval (frequency) for capturing the image information of the flock image obtained via the ITV 30 and the ITV controller 32 shown in the figure is controlled. Next, the A / D conversion circuit 101
Receives the analog signal of the obtained luminance information, for example, the image signal of FIG. 4, and finally converts the signal into a digital signal.

FIG. 4 shows an image surface of the flock group recognized by the ITV 30. Since the flock group is a grayscale image, the boundary of the flock 12 with water is not clear, but for simplicity, only the ring part of the flock group is shown. The brightness level of the flocs is high and white, while the background is the back screen 2
Since 4 is arranged, the brightness level of water is low and black.

The converted digital signal is binarized in the binarization circuit 103 based on the threshold value specified by the threshold value input circuit 102.

For example, FIG. 5 shows a case where the luminance level distribution is displayed by scanning along the line AA 'on the screen of FIG. Here, the brightness level is displayed in 8 bits (256 levels),
Brightness is low in the upper direction of the vertical axis, while high in the lower direction.
Since the flock 12 is white, the brightness is high. That is, the part which becomes a valley below represents a floc.

In this luminance distribution, each pixel is binarized by the binarizing circuit 103 based on the threshold value designated by the threshold value input circuit 102, for example, the luminance designated by the BB ′ line. The threshold value specified by the threshold value input circuit 102 is kept constant under constant illuminance, but can be operated by the operator.

In the binarization circuit 103, pixels having a brightness level higher than the threshold value are set to “1”, while pixels having a brightness level higher than a predetermined value are set to “0”. Then, as shown in FIG.
The part corresponding to the flock becomes "1", and the part corresponding to water becomes "0".

FIG. 7 shows an example of the result of recognizing the flocs in this way.

When the flocs are recognized, the representative particle size of the flocs is calculated next, but each floc is numbered before that. That is, the labeling circuit 104 independently recognizes each flock and assigns a number to each flock. Then, the particle size measuring circuit 106 sets the flocs in order of the numbers.
Calculate the representative particle size in.

As the representative particle diameter, there are a unidirectional diameter D _c , a maximum D _max , a minimum diameter D _min , an area circle equivalent diameter D _{cir, an} equivalent peripheral major axis D _c1, and the like. The grain size calculation mode designating circuit 105 designates a representative grain size to be adopted from these representative grain sizes. In accordance with the standard of the representative particle size designated in this way, the particle size is calculated for each floc.

In the particle size comparison circuit 107, the particle sizes of the flocs are compared, and the number of flocs having each particle size is stored in a corresponding storage location, that is, the number storage circuits 108A, 108B, 108.
The numbers are stored in C, ... 108Z. Since the flock image is binarized, the minimum unit for measuring the particle size is one pixel which is predetermined by the CRT. Therefore, if the number of flocs corresponding to each particle size is N _i ,
For example, the number storage circuit 108A stores the number N _{1 of} flocs whose particle size corresponds to one pixel, and the number storage circuit 10A
In 8B, the number N _{2 of} flocs whose grain size corresponds to two pixels is stored. Although the minimum unit of particle size is 1 pixel, the minimum unit of 2 pixels is determined only by the purpose of the flock monitoring, the condition of the flock formation pond, and the image captured by the TV camera.

By the way, in order to obtain the particle size distribution of flocs with high accuracy,
It is better to capture the screen multiple times and obtain the particle size distribution from a lot of information. The number of times of recognition of the flock image is designated in the recognition number control circuit 109. If the number is less than this number, the process returns to the recognition timing control circuit 100. The recognition timing control circuit captures an image at a specified timing and repeats the operation described above to calculate the particle size of flocs, and stores the number of each floc in the number storage circuits 108A, 108B, 108C, ... 108Z. To do.

On the other hand, when the number of times of recognition is reached by the recognition number control circuit 109, the particle size distribution calculation display circuit 110, based on the values of the number storage circuits 108A, 108B, 108C, ... Is calculated and the result is output.

Next, the automatic threshold value determination, which is a feature of the present invention, will be described.

The automatic threshold determination is performed by operating the image recognition processing unit 50 described above in the automatic threshold determination mode. In order to determine the automatic threshold value, the threshold value is successively changed with respect to the reference image captured by the TV camera 30, and the relationship between the floc particle size and the number is calculated for each threshold value. Next, this relationship is searched for each threshold value, and threshold value determination processing is performed to determine the optimum threshold value.

In this automatic determination operation, the threshold value is changed by the automatic threshold value determination circuit 90, and the binary value is changed through the input circuit 102.
Send to circuit 103. The relationship between the floc particle size and the number for each threshold value is stored in the memory 108. The search of the contents of the memory 108 and the optimum threshold value determining process are performed by the automatic threshold value determining circuit 9
0 does.

The threshold value determination method will be described more specifically.

To simplify the explanation, consider the case where the grayscale image shown in FIG. 8 is binarized. The flock is almost spherical,
Therefore, the center of the flock has the highest brightness level, and the brightness distribution spreads from the center like a contour line, and the brightness level becomes lower toward the outside. To simplify the description, consider eight equal-level lines 201 to 208, the equal-level line 201 as level 4, the equal-level line 202 as level 5, and the equal-level line 203.
Let level 6 be the level 6, the level line 204 be level 7, the level line 205 be level 8, the level line 206 be level 9, the level line 207 be level 10, and the level line 208 be level 2. C and D are overlapped by flock. Then 2
When the threshold for binarization is set to 1 and the one having a brightness level of 1 or more is considered to be a floc, the grayscale image 200 of FIG.
It is binarized like a binary image 210 in the figure. The shaded portion is the portion that is determined to be white when binarized. Next, when the threshold for binarization is selected as 2, the grayscale image 200 in FIG. 8 is binarized like the binary image 210 in FIG. When the threshold value for binarization is selected as 3, the grayscale image 200 in FIG. 8 is binarized like the binary image 211 in FIG. 10, and the distinction between backboard (water) and floc appears. Next, even when the binarization threshold value is selected as 4, the grayscale image 200 in FIG.
It is binarized like a binary image 211 in FIG. Next, when the binarization threshold is selected as 5, the grayscale image 200 of FIG.
It is binarized like the binary image 212 in FIG. Then 2
When the threshold for binarization is selected as 6, the grayscale image 200 of FIG.
Is binarized like a binary image 213 in FIG. Next, when the threshold for binarization is selected as 7, the grayscale image 2 of FIG.
00 is binarized like the binary image 214 of FIG. 13,
The number of objects is two. Next, when the threshold for binarization is selected as 8, the grayscale image 200 in FIG. 8 is binarized like the binary image 215 in FIG. Next, the threshold for binarization is set to 9
When selected as, the grayscale image 200 of FIG. 8 is binarized like the binary image 216 of FIG. Next, when the threshold for binarization is selected as 10, the grayscale image 200 in FIG.
It is binarized like the binary image 217 of FIG. 6, and the number of flocs is reduced to one. Next, when the threshold value for binarization is selected to be 11 or more, the grayscale image 200 in FIG.
The value image 218 is binarized, and the number of flocs becomes zero. As described above, when the threshold for binarization is changed,
The number of objects (flocks) and the average area of the objects change. First, paying attention to the number of objects, the number of objects is
There is a threshold range in which the number increases when the threshold is increased, but decreases when the threshold is exceeded and the number of objects is maximum. Looking at the average area of the object, the average area decreases as the threshold value is increased. Further, in the binary images 210 to 218 in FIGS. 9 to 17, when the binary image which is accurate as the information having the most flock is considered,
It is considered that the binary image 214 in FIG. 13 in which the average area is maximum in the binary screen in which the number of flock is 2 and the number of flock is 2 is accurate.

Therefore, the threshold for binarization is sequentially changed through the input circuit 102, the number of flocs is calculated for each threshold, the area of each floc is calculated, and the results are stored in the memory 108 to determine the automatic threshold. The circuit 90 searches the contents of the memory 108, selects a threshold value when the number of flocs is the maximum and the floc average area is maximum, and determines this as the official threshold value.

The determination mechanism will be described with reference to FIGS. 18 (a) to 18 (c). FIG. 18 (a) is a diagram in which both the horizontal axis and the vertical axis are displayed by the threshold value V _th , and FIG. 18 (b) shows the threshold value V _{th on the} horizontal axis and the number M of flock on the vertical axis. In FIG. _18C , the horizontal axis shows the threshold value V _th and the vertical axis shows the average flock area S. Threshold V _th
Takes a value of 0 to 127. In FIG. 18B, the section A is the section in which the number of flocs is the maximum, and in this section A, the threshold value V ₁ ≦ V _th ≦ V ₂ is set, and the same maximum number continues. This is the section that occurs. On the other hand, in FIG. 18C, the average area S of flocs for each threshold value V _th is produced and displayed. This area curve S and V ₁ , V ₂
Let S ₁ and S ₂ be the intersection points with S ₁ > S ₂ and S _1
1 is the maximum average area in the section A.

Therefore, the optimum threshold value V _{th to be obtained} is V _th = V ₁ (=
P ₁ ). After the optimum threshold value is determined, the automatic threshold value determination mode is switched to the flock recognition mode, and the flock recognition is performed according to the process shown in FIG.

The automatic threshold value determination mode may be operated in advance, or may be appropriately operated during the process of forming flocs.

The automatic threshold value determining circuit 90 itself includes circuits 100 to 100.
You may have 108. In that case, the threshold can be automatically determined independently.

〔The invention's effect〕

According to the present invention, by appropriately determining the threshold value, it is possible to prevent the deviation of the threshold value setting of the binarization due to the fluctuation of the ambient light of the underwater camera, the fluctuation of the turbidity, etc. Allows unattended continuous processing.

[Brief description of drawings]

FIG. 1 is an embodiment diagram of the present invention, FIG. 2 is a configuration diagram of a flock formation pond which is an object of the present invention, FIG. 3 is an overall configuration diagram of the present invention for a flock formation pond, and FIG. 4 is a flock image. 5 and 6 are explanatory diagrams of binarization, FIG. 7 is an example of binary image, FIG. 8 is a relational diagram between flock and contour line level, and FIGS. 9 to 17 are threshold values. Binary image example by changing,
FIG. 18 is an explanatory diagram of determining the optimum threshold value. 50 ... Image recognition processing unit, 60 ... Image recognition control device, 90 ... Automatic threshold value determination circuit.

Claims (1)

[Claims] 1. A flock image recognition device comprising: an image pickup means provided in a flock formation pond for picking up a flock group; and an image processing means for binarizing a picked-up image by an optimum threshold value and performing flock recognition from the binarized image. A threshold determining means is provided for changing the threshold for binarization to obtain a threshold that maximizes the number of flocs in the binarized image and maximizes the average area of the flocs, and sets the threshold as the optimum threshold. Flock image recognition device.