JPH0621851B2 - Flock image recognition device - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to a floc image recognition device for measuring the particle size of flocs and its distribution in a flocks forming pond (mixing pond) of a water purification plant using an image processing technique.

[Background of the Invention]

Since the turbidity particle size of the raw water is small in the water treatment plant, it is a process of aggregating these to form a flocculation pond and allowing the flocs to settle. Therefore, it is essential to monitor the flock in the flock formation pond (mixing pond).

Conventionally, the monitoring of the flocks has been performed visually by the maintenance manager of the water purification plant several times a day. However, since it depends on visual observation, there is a problem in that the judgment standard is subjective and qualitative, and the monitoring result is difficult to be reflected in the driving operation. In addition, since the monitoring frequency is discontinuous, it is unavoidable that measures for cohesive failure will be delayed and trouble will increase.

In order to solve this problem, a method of monitoring the flocks in the flocks forming pond using an industrial television camera (ITV) has been adopted recently. However, even in this case, monitoring is still subjective and discontinuous because it depends on human vision. Therefore, JP-A-54-1
As described in Japanese Patent No. 43296, there has been proposed a method of extracting the shape of a block as an electric signal using a photoelectric conversion device. In order to put this method into practical use, put an ITV (industrial TV camera) in an airtight container, install it in the water of the flock forming pond, take a picture through the observation window provided in the airtight container, and capture the flock image in the image processing device. It is common to In this case, if dust or flock adheres to the observation window and the backboard attached to the opposite side of the observation window, these immovable objects are caught by the ITV and treated as an analog signal of the same level as the flock, so the flock formation situation in the flock formation pond. It becomes a big error in knowing. For this reason, it is necessary to quickly know and remove the objects adhering to the observation window and the backboard.

[Object of the Invention]

An object of the present invention is to provide a floc image recognition device capable of detecting dirt on an observation window online when recognizing the particle size or distribution of floc group by image processing.

[Outline of Invention]

A feature of the present invention is that the number of pixels showing a binary signal state that is the same as the number of times of image processing does not change between the number of pixels by the previous image processing and the number of continuous image processing of two or more times. Sometimes, it is detected as dirt on the observation window.

Example of Invention

 FIG. 1 shows an embodiment of the present invention.

In FIG. 1, an industrial television camera (ITV) is housed in an airtight container 20 and installed in the water of the flocks forming pond 10. The ITV 30 fixed in the airtight container 20 is located in the flocks forming pond 10 through the observation window 21 made of a transparent material such as glass by the close-up lens 31.
Magnify the second image. A wiper 22 is provided which is regularly driven by a wiper drive device 23 to remove dirt on the surface of the observation window 21. A back screen 24 is installed at a position facing the close-up lens 31.
The back screen 24 is provided in order to accurately recognize the block group 12 with high contrast, and is attached to the airtight container 20 by back screen fixing tools 24A and 24B. The black screen 24 is preferably black because the color of the block group 12 is white. A light blocking cover 42 is fixed to the block formation pond 10. The flock formation pond 10 is normally open to the atmosphere.
Therefore, the amount of light incident on the flock forming pond 10 changes with the passage of time and is strongly influenced by the weather. The light shielding cover 42 is provided to eliminate the influence of changes in incident light. The lighting device 40 illuminates the imaging area of the ITV 30.
The illuminance of the lighting device 40 is adjusted by the illuminance adjusting device 41. The change in illuminance in the shooting area strongly affects the image recognition accuracy of the flocks. For example, if the illuminance is low, the block is recognized as small, and conversely, if the illuminance is high, the block is recognized as large. In order to remove this effect, it is desirable not to be affected by changes in illuminance as a natural phenomenon. In this embodiment, a light-shielding cover 42 is provided to darken the surroundings, and the illumination device 40 provides a constant illuminance. A flow velocity limiting plate 25 is provided on the upper portion between the airtight container 20 and the back screen.
A, 25B and 25C are provided. Flow rate limiting plate 25
A, 25B, and 25C are provided to slow down the moving speed of the block near the observation window 21 and obtain a stable image. The block image information captured by ITV30 is I
It is input to the image processing device 50 via the TV controller 32. The image processing device 50 performs various calculations such as the particle size distribution of the flocks in order to extract information useful for water quality management of the water purification plant from the obtained image information.

FIG. 2 shows an example of the image processing device 50.

In FIG. 2, the timing control circuit 100 is ITV3.
0 and the timing of fetching the block image information obtained via the ITV controller 32 is controlled. The timing operates when the process is activated at a specified time interval and when the recognition number control circuit 109 activates. Next, the A / D conversion circuit 101 receives the analog signal of the obtained brightness information, for example, the screen signal of FIG. 4, and converts it into a digital signal of this image signal. The converted digital signal is binarized in the binarization circuit 103 based on the threshold value specified by the threshold value setting circuit 102. For example, FIG. 5 shows a case where the distribution of the brightness level is displayed by scanning the line AA 'in FIG. Here, the brightness level is displayed at 8 bits (256 levels), and the brightness is lower in the upper direction and higher in the lower direction. Since the block 12 is white, the brightness is high. Fifth
The valley at the bottom of the figure represents the flock. Each pixel is binarized by the binarization circuit 103 based on the threshold value designated by the threshold value setting circuit 102, for example, the luminance designated by BB ′. The threshold value specified by the threshold value setting circuit 102 is kept constant under constant illuminance, but can be changed by the operator. In the binarization circuit 103, pixels having a brightness level higher than the threshold value are set to the “1” level, while pixels having other brightness levels are set to the “0” level. As shown in FIG. 6, the binarized signal has a "1" level at the portion corresponding to the block and a "0" level at the portion corresponding to water. FIG. 7 shows an example of the result of the recognition of the block in this manner.

The binary image processed by the binarization circuit 103 is input to the window stain detection device 7 and detects dirt due to adhesion of dust on the observation window 21. Details of the window stain detection device 7 will be described later.

On the other hand, the labeling circuit 104 assigns a number to each block in order to calculate the particle size of each block. That is,
Individually recognize each flock and assign a number to each flock. The particle size measuring circuit 105 calculates the representative particle size of each block in numerical order. The representative particle size is obtained by counting the number of pixels in the area of the numbered block,
Assuming that the shape of the flock is a sphere, find the area circle equivalent diameter. The particle size display circuit 110 outputs the representative particle size calculated by the particle size measurement circuit 105.

By the way, when performing statistical processing based on the representative particle size,
It is desirable to capture the screen multiple times and obtain a lot of information to obtain it. The recognition number control circuit 109 specifies the number of times of recognizing a block image. If the number is less than this number, the recognition timing control circuit is activated and the operations described above are repeated.

Next, the window stain detection device 7 for the observation window will be described.

FIG. 3 shows a detection flow by the window stain detection device 7. The abnormal image is detected by the binarization circuit 103.
Although this is performed by processing the value image, for the sake of simplicity, consider the case where the lines shown in the pixel configuration 200A in FIG. 8 have 6 pixels in the vertical direction and 6 pixels in the horizontal direction. In FIG. 8, the storage memory A _XY (A ₁₁ , A ₁₂ ...
A ₆₆ ) is provided. This storage memory is called a binary pattern integration memory.

First, in step S1, the number N of screens used for detection
Memorize The number of screens N is calculated as N = N + 1, and N is cleared to zero when the system is started up. Next, in step S2, after the number of processed screens N = 1, the process proceeds to step S3 to clear the binary pattern integration memory A _XY to zero, and then proceeds to step S4. When the number of processed screens N ≠ 1, the process directly proceeds to step S4. In step S4,
In the binary image, 1 is added to the binary pattern integration memory corresponding to the "1" level pixel. (A _XY = A _XY +1). Next, in step S5, the number N of screens obtained in step S1
The number of binary pattern integration memories that match the value of is extracted and stored in the memory M. If the number of pixels of the true dust portion is P, it can be considered that M ≦ P. In step S6, M =
In the case of 0, it is determined that there is no stain, and step S1
The process proceeds to 3 and the screen number N is cleared to zero, and the process ends.
If M ≠ 0 is determined in step S6, step S7
Move to. In step S7, when the number of screens N = 1, the process proceeds to step S12, the contents of the memory M are stored in the memory M _N-1 for the next comparison, and the process is terminated.
If the number of screens N ≠ 1 in step S7, the process proceeds to step S8. In step S8, the coincidence signal is reset, the memories M and M _N-1 are compared, and if they do not coincide (M ≠ M _N-1 ), the process proceeds to step S10. In case of coincidence (M = M _N-1 ), the process proceeds to step S9 to set the coincidence signal and then proceeds to step S10. In step S10, it is determined whether the coincidence signal has been set twice in succession. If the coincidence signal has been set twice in succession, the process proceeds to step S11, an alarm is output, the number N of screens is cleared to zero, and the process is terminated. The reason why the alarm is output under the condition that it is set twice consecutively may be to judge that dust is detected when the coincidence signal is set, but when P <M, M = M
_Since it may be _N-1 , it is judged that dust has been detected when the coincidence signals occur consecutively. Further, the false alarm is less likely to be output when the alarm is output when it is set three times consecutively than when it is set twice consecutively. In step S10, if the data has not been set twice consecutively, the contents of the memory M are stored in the memory M _N-1 and the process is terminated.

The above will be described with reference to the drawings.

In FIG. 9, 201 is a dirty portion of the observation window. Considering that dirt is attached here, consider a case where images are captured in the order of 202A in FIG. 10 (a) to 206A in FIG. 14 (a). Figure 10 (a)
202A is a binary image. First, the number of screens N = 1 is set in step S1, the binary pattern integration memory A _XY is cleared to zero in step S3, and then step S4 is executed. Like Then, in step S4, the memory M = 6 is stored. Since N = 1, the comparison process with the previous value is not performed, and the memory M is read in step S12.
The contents of the above are stored in the memory M _N-1 . M _N-1 = 6.
Next, when 203A in FIG. 11A is processed, the number of screens N
= 2, the binary pattern integration memory is 2 in FIG. 11 (b).
03B, and the memory M = 4 is obtained. The process proceeds to step S8, and from M ≠ M _N-1 (4 ≠ 6), step S10
Since the coincidence signal is not set twice in succession, the process proceeds to step S12 and the memory M _N-1 = 4 is stored. Then the first
When 204A in FIG. 2A is processed, the number of screens N = 3, 2
The value pattern integrating memory is as shown in 204B of FIG. 12 (b), and the memory M = 4 is required. At step S8, the coincidence signal is set at M9 from M = M _N-1 (4 = 4). The number of pixels of true dirt is 2, whereas M =
Therefore, when the coincidence signal is detected, the number of miss alarms increases when the alarm is immediately issued. Since the match signal may not be set by processing the next screen,
As for the alarm output determination, the number of error alarms is smaller when the alarm output is determined when the coincidence signals are continuously set. Since the coincidence signal is not set twice consecutively in step S10, the process proceeds to step S12 and the previous value M _N-1 = 4 is stored.
Next, 205A of FIG. 13 (a) and 20 of FIG. 14 (a).
When 6A is processed in the same manner, 209 in FIG.
When A is processed, M = 2 and the number of pixels of true dirt is 2
Since it converges on, no matter what screen (including the dirt part) is processed thereafter, M = M _N-1 = 2 and the coincidence signal is continuously set, and an alarm is output. .

〔The invention's effect〕

As described above, according to the present invention, the dirt on the observation window of the water purification block image recognition device can be detected online only by the image processing device, so that regular maintenance can be improved and highly reliable image processing can be performed unattended. It is possible to continuously process the.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a detailed block diagram of an example of an image processing apparatus, FIG. 3 is a flow chart of window stain detection, and FIGS. 7 is a binary image diagram obtained by carrying out the present invention, FIG. 8 is a diagram showing a screen structure, FIG. 9 is a diagram showing an abnormal pattern, and FIGS. It is a figure which shows a value image and the binary pattern integration table in a process. 7: Window stain detector, 10: Flot formation pond, 12
...... Flock group, 20 ...... airtight container, 21 ...... observation window,
22 ... Wiper, 23 ... Wiper drive device, 24 ...
... back screen, 25A, 25B, 25C ... flow rate limiting plate, 30 ... industrial TV camera (ITV), 31 ... close-up lens, 32 ... ITV controller, 40 ... illumination device, 41 ... illumination adjustment device , 42 ... Shading plate, 50 ...
Image processing device.

Front page continued (72) Inventor Mikio Yoda 5-2-1 Omika-cho, Hitachi City, Ibaraki Hitachi Ltd. Omika Plant, Ltd. (72) Inventor Shunji Mori 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture Hitachi Ltd. Omika factory

Claims (1)

[Claims] 1. An industrial television camera which is disposed in the water of a flock forming pond and repeatedly photographs flock through an observation window; a back screen which is disposed opposite to a close-up lens of the industrial television camera; An illumination means for illuminating a photographing area of the television camera, and an image processing means for dividing an image photographed by the industrial television camera into a flock and a background for each pixel and converting into a binarized signal to perform image processing. The image processing means, when the number of pixels showing a binary signal state that is a floc the same number of times as the number of times of image processing does not change continuously from the number of images by the previous image processing twice or more. A flock image recognition device having a window stain detecting means for determining that the observation window is stained.