JP2005150840A - Water quality monitoring system and fish image recognition method used for same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enhance the accuracy of monitoring by eliminating misjudgment caused by an image other than a fish and an active fish without moving. <P>SOLUTION: A water quality monitoring system comprises a water tank 1 in which fishes are put, a camera 2 for imaging the water tank 1 at a constant time interval and a computer 3 for applying image processing to an image photographed by the camera 2, and the computer 3 includes a recognition means for recognizing the fish image and a judgment means for judging the suitability of water quality on the basis of the recognized fish image. The recognition means performs a step of selecting a group of a plurality of images captured from the camera 2 to the computer 3, a step of applying average processing to the luminance of each pixel in the selected image group to generate a background image; a step of selecting an object image which is monitored and judged, a step of subtracting the background image from the object image to obtain a difference image, and a step of recognizing the fish image of activating fishes on the basis of the difference image. A deteriorated level of the water quality is judged on the basis of detailed motions of the fishes after performing the above steps. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a water quality monitoring device and a fish image recognition method used therefor. More specifically, small fish (eg, medaka) that are sensitive to toxic substances as living organisms are placed in a water tank, and the captured small fish images are processed by a computer to automatically track the behavior of each small fish in real time. The present invention relates to an apparatus and method for detecting water quality abnormality.

In recent years, there has been a growing interest in protecting safe water from environmental degradation caused by terrorism and illegal dumping.
When inspecting the water quality of rivers and ponds, an unspecified number of toxic substances cannot be detected by a water quality inspection device using an industrial instrument. In addition, depending on the type of toxic substance, it takes time for pretreatment and component analysis, making it difficult to respond quickly.
Therefore, water-related facilities such as water purification plants often detect abnormalities in water quality using not only industrial instruments but also fish and other organisms. In this case, most of them are observed visually, but there is a problem that it is not practical because it requires a monitoring person to monitor the fish 24 hours a day to determine the subtle behavior of the fish.

For this reason, an apparatus for automatically detecting an abnormality in water quality from the behavior of fish or the like, action potential and heart rate has been developed. However, with this device, various sensors must be attached to measure the action potential, heart rate, etc. of the fish, and there are problems in terms of the complexity of maintenance and the frequent occurrence of false alarms.
On the other hand, high-speed and multi-functional image processing technology is progressing due to improved computer performance, and if this technology is used to monitor the behavior of fish in water, high-precision water quality monitoring can be expected. Type water quality monitoring devices have been developed (for example, Patent Documents 1 and 2).
The fish image recognition method used in the above-described conventional technique recognizes the bright spot of the captured image as a fish by photographing the water tank containing the fish at regular time intervals and binarizing the image. At the same time, in order to determine the presence or absence of activity of the fish, it is a method of calculating and monitoring the movement distance of the fish generated between the images before and after the time.

However, this fish image recognition method has the following problems.
(1) There is a possibility of recognizing as fish even objects floating in the aquarium, sediments, and reflection on the glass surface.
(2) Since the monitoring target is a moving distance, a fish that is active without moving may be mistaken for a stop of activity.
(3) When muddy water is mixed in the source water and the turbidity cannot be completely removed by a filter or the like, if the figure of the fish is hidden due to the turbidity and it is difficult to recognize the fish image, normal water quality determination cannot be performed. As a countermeasure, a method of installing a turbidimeter in the previous stage and automatically stopping water intake when the turbidity increases to an unrecognizable level can be considered, but the apparatus becomes large and complicated and expensive.
(4) If a small (young) fish is introduced, it is possible to monitor for a long time, but if the fish recognition accuracy decreases, and if a large (old) fish is introduced, the fish recognition accuracy increases, but natural death Is not suitable for monitoring. For this reason, the actual age of the moon can be determined to some extent from the size, so when the fish is thrown in, the size is chosen and thrown, but it is difficult to judge whether the thrown fish is suitable for monitoring .
(5) Depending on the fish selected, there may be ferocious fish that chase other fish, making natural death more frequent and not suitable for monitoring. The presence of ferocious fish is difficult to notice unless the fish movement is constantly monitored.

Japanese Patent Publication No. 5-31941 Japanese Patent Publication No.7-85082

In view of the above circumstances, the first object of the present invention is to improve the monitoring accuracy by eliminating misjudgments caused by images other than fish and fish that are active without moving.
A second object is to measure the turbidity in the water tank so that the turbidity can be managed.
A third object is to make it possible to determine whether a fish is suitable for water quality monitoring.

The water quality monitoring device of the first invention comprises a water tank containing a water quality monitoring fish, a camera for photographing the water tank at regular time intervals, and a computer for judging the water quality based on the images taken by the camera, The computer has image recognition means for recognizing a fish image, and water quality determination means for determining the suitability of water quality from the recognized fish image, and the image recognition means is taken into the computer from the camera. Selecting a plurality of image groups, creating a background image by averaging the luminance of each pixel in the selected image group, selecting a target image to be monitored, and the target And subtracting the background image from the image to obtain a difference image, and recognizing a fish image of an active fish based on the difference image.
A water quality monitoring device according to a second aspect of the present invention is the water quality monitoring device according to the first aspect of the present invention, comprising a illuminating lamp for illuminating the aquarium, wherein the computer comprises turbidity judgment means for water in the aquarium, The degree determining means includes a step of detecting a luminance ratio between a water area close to and far from the illuminating lamp in the water tank, and a step of determining whether the luminance ratio is equal to or higher than a reference value. And
The water quality monitoring device according to a third aspect of the present invention is the water quality monitoring device according to the first aspect, wherein the computer comprises suitable fish judgment means for judging whether or not the fish in the aquarium is suitable for water quality monitoring. The step of measuring the projected area of the fish when the fish is swimming near the center of the image in which the camera is photographing the aquarium, and the measured projected area is less than or equal to the lower limit value of the reference value or the upper limit value It consists of the step which judges whether it is above.
The water quality monitoring device according to a fourth aspect of the present invention is the water quality monitoring device according to the first aspect, wherein the computer includes suitable fish determination means for determining whether or not the fish in the aquarium is suitable for water quality monitoring. The method includes a step of measuring acceleration during swimming and a step of determining whether or not a sum of absolute values of the measured accelerations is equal to or greater than a reference value.
The fish image recognition method of the fifth invention is a method for recognizing a fish image by photographing a water tank containing a water quality monitoring fish with a camera at regular time intervals, and processing the obtained image with a computer. Selecting a plurality of image groups captured from the camera to the computer, creating a background image by averaging the luminance of each pixel in the selected image group, and monitoring. Selecting a target image to be determined, subtracting the background image from the target image to obtain a difference image, and recognizing a fish image of an active fish based on the difference image. It is characterized by that.

According to the first invention, when recognizing a fish image, a background image is created by averaging a plurality of captured images, and the activity is stopped by taking a difference from the target image to be monitored and determined. It is possible to obtain a differential image that eliminates the influence of the reflected fish, floating objects, sediment, and reflection on the glass surface. By performing image processing using this difference image, it becomes easier to grasp even small movements as image changes, so if you are doing fine operations such as changing the tilt of the body even with fish that are not moving, Can be recognized as active. Thus, since the fine movement of the fish can be monitored, water quality monitoring based on the behavior of the fish can be performed with high accuracy.
According to the second invention, when the turbid water in the aquarium is illuminated from one direction, irregular reflection occurs in the water area close to the illumination and brightens, and this irregular reflection makes the light far away from the illumination and darkens. . Since this tendency becomes more prominent as the turbidity is higher, the turbidity of water in the water tank can be indirectly measured by comparing the luminance ratio with a reference value. In this way, if the turbidity where the fish image cannot be recognized normally is set as the reference value, the turbidity warning can be issued before the warning due to the inability to recognize the fish.
According to the third invention, since the projected area is measured while swimming in the vicinity of the center of the image for each inserted fish, the direction of the body relative to the camera and the distance from the camera can be made substantially constant. . And it can detect young fish with small projected area, that is, small body and poor recognition rate, and old fish with large projected area, that is, large body and slow movement. Can improve the monitoring accuracy.
According to the fourth invention, when the acceleration of the fish is measured and it is determined whether or not the sum of the absolute values of the acceleration over a certain time is greater than or equal to the reference value, the fish whose acceleration changes frequently, that is, other fish are more than necessary. You can find ferocious fish that chase you. Thus, the water quality monitoring accuracy can be improved by exchanging fish that are not suitable for monitoring.
According to the fifth invention, when recognizing the fish image, the activity is stopped by averaging a plurality of captured images to create a background image and taking a difference from the target image to be monitored and judged. It is possible to obtain a differential image that eliminates the effects of fish, floating objects, sediment, and reflection on the glass surface. By performing image processing using the difference image, it becomes easier to grasp even small movements as image changes, so if you are doing fine operations such as changing the tilt of the body even with fish that are not moving, Can be recognized as active. Since it is possible to monitor fine movements of fish, water quality monitoring based on fish behavior can be performed with high accuracy.

Next, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 shows a water quality monitoring apparatus according to an embodiment of the present invention, in which 1 is a water tank, 2 is a camera, and 3 is a computer. The water quality monitoring apparatus of the present embodiment can be configured by the water tank 1, the camera 2, and the computer 3, and is characterized by a simple configuration that is completed when various programs for executing water quality monitoring are incorporated into the computer 3.

In the present embodiment, the water tank 1 and the computer 3 can be installed in different places, respectively, and can be monitored from a remote place. It is also possible to transfer collective monitoring and alarms of a plurality of water tanks 1 to a mobile phone or the like.
FIG. 2 shows an example of remote monitoring. Arbitrary communication means such as video cable, wireless LAN or wired LAN, wireless or wired dedicated line, public line can be used between the camera 2 and the computer 3, and the distance between the water tanks 1A to 1D and the computer 3 is as follows. From a few meters to a few tens of kilometers is practical, but remote monitoring is possible with existing technology even at distances beyond that.

The water tank 1 may be of an appropriate size, but it is preferable that the monitoring area 12 is narrowly limited by inserting a partition 11 or the like. The fish to be placed in the aquarium 1 is preferably a small fish that is easy to handle and highly sensitive to harmful substances, but medaka which is particularly easily available is suitable. Medaka has the characteristics that it reacts when it is mixed with toxic substances, such as “rage”, “gather in a specific place”, “slow movement”, “stop”, etc. About 200 cases such as “cyan” and “arsenic” Highly reliable judgment results are obtained for poisonous substances.
Further, it is preferable to place about 10 medaka f in the monitoring area 12 in the water tank 1 because the individual medaka f can be specified or the monitoring accuracy can be improved and the image processing does not become more complicated than necessary. Although a water tank of a size corresponding to the monitoring area 12 may be used, the water tank 1 having a large non-monitoring area 13 as shown is more convenient for practical work such as replacing the medaka f. is there.
The water tank 1 is supplied with water for water quality inspection through a filter, and drains the water tank after running water.

  An illuminating lamp 4 for irradiating light is provided above the water tank 1. When the illumination lamp 4 is irradiated, the fish appears as a bright spot, and this bright spot is used for recognition of the fish image.

The camera 2 is not particularly limited as long as it can shoot fish images, but any camera can be used, but a camera that can stably shoot for a long period of time is preferable.
Shooting with the camera 2 is preferably performed continuously at regular intervals. It is preferable that the time interval of image processing is short because the behavior of the medaka can be traced finely. However, if the time interval is too short, the image processing cannot catch up and the monitoring accuracy cannot be improved much. The preferred range is about 0.033 to 0.5 seconds, but in the present invention, the intervals are 0.1 seconds.
Therefore, the camera 2 continuously shoots the water tank 1 at intervals of 0.1 seconds, sends image data to the computer 3, and the computer 3 measures the movement of the medaka f about every 0.1 seconds and performs high-speed calculation processing as image data. For this reason, water quality can be determined in real time.
In addition, the number of medaka f serving as sensors is about 10 as described above, and is measured every about 0.1 second, so that a fine movement can be detected and an abnormality in water quality can be determined with high accuracy.

The computer 3 has a program for recognizing a fish image based on an image taken from the camera 2 and judging the water quality, and a program for judging whether the water in the aquarium 1 is suitable for turbidity or monitoring of the water. Is installed. The image recognition means, water quality judgment means, turbidity judgment means, and suitable fish judgment means referred to in the claims are constituted by these programs. It should be noted that these image processing and determination can be handled by a personal computer commercially available.
The computer 3 also has a function of issuing a necessary alarm based on the result of executing each of the determination means. The method of issuing an alarm is arbitrary, and various methods can be adopted. In addition, the alarm can be displayed on the monitor 3A of the computer 3 or in addition to the necessary communication means by an appropriate communication means, or an alarm sound can be generated. You can adopt a method to sound. A monitor screen such as an alarm to be displayed on the monitor 3A is arbitrary, but an example thereof will be described later.

Next, a water quality judgment method executed by the water quality monitoring device will be described with reference to FIG.
This water quality judgment method is roughly divided into an image recognition step and a water quality judgment step based on the image recognition step. These methods are executed by the fish image recognition means (program) and the water quality judgment means (program) installed in the computer 3 as described above.

Next, the recognition method by the fish image recognition means will be described in the order of each step.
1 Multiple Image Group Selection Step (S1)
In the water quality monitoring apparatus of the present embodiment, 20 images (images of about 2 seconds) stored in the memory in order from the camera 2 to the computer 3 are stored in the memory, and the new image is overwritten on the oldest image. 20 images are constantly updated.
These 20 past images are selected for creating a background image, but a smaller number may be selected as appropriate.

2. Background image creation step (S2)
Next, using the 20 images, the luminance of each pixel is averaged to create a background image.
The technique for averaging the luminance is as follows.
In the digital image, 256 levels of gradation (8 bits) are given for each color of RGB (red, green, blue), so there are a total of 16.77 million patterns such as R0G0B0 = black, R255G0B0 = red, R255G255B255 = white. Color display is possible. In the case of image processing, (R × 0.33 + G × 0.33 + B × 0.33) is calculated from 256 RGB gradations, and a new 256 gradation monochrome image is created. At this time, the coefficient 0.33 is not fixed, and the fish color may be emphasized and the recognition rate may be set high. The 256 gray scale colors are used as the “brightness” in the averaging process. That is, assuming that the luminance K of the pixels at the same position in the past 20 image groups is (K1, K2,..., K20) in the newest order,
(Kb = K1 × a1 + K2 × a2 +... + K20 × a20) (1)
However, (a1 + a2 +... + A20 = 1)
Can be averaged, and the value of the luminance Kb of the pixels of the background image can be given. The coefficient a can be set freely, but normally, a1 ≧ a2 ≧... ≧ a20.

3. Target image selection step (S3)
A target image to be monitored is selected. Usually, the latest image is selected, but it is also possible to select a past image.

4. Difference image creation step (S4)
A background image is subtracted from the target image to obtain a difference image.
Since both the background image and the target image (current image) are multi-valued images, the subtraction process is performed with the multi-valued image as it is. The method for obtaining the difference image is as follows.
For the sake of easy understanding, a binary image will be described. If the bright area including the fish image is “1” and the dark area is “0”, the difference is calculated as shown in the table below.
Target image-Background image = Difference image
1 1 0
1 0 1 *
0 1 -1
0 0 0
Of the above, only the “1” area of the difference image is used for fish image recognition. The recognition procedure is as follows.
1) The area “1” is enlarged (when there is a divided “1” area, the area is enlarged to a level where the adjacent areas become one).
2) The area “1” is reduced to the original size (the combined area is reduced as a whole).
3) Delete areas larger or smaller than the set fish image size.
4) Recognize the remaining “1” area (* in the table above) as a fish image, and detect the coordinates of each fish and the movement status from the previous time, using the number of “1” areas as the number of fish.
5) To increase recognition accuracy, use the number of “1” areas that have been tracked 20 times (about 2 seconds) as “number of activities”.
As described above, since a multi-valued image is used in the present embodiment, when applying to the above procedures 1) to 5), if processing is performed using an appropriate threshold value, fish image recognition is performed in the same manner as described above. Can be implemented.
In short, obtaining the above difference image means
B) Objects that exist in both the target image (latest image) and the background image are removed. → In other words, fish and stationary objects that are not moving are removed. B) As a result, only the target image remains. All that is left is the moving fish, including those that have changed position and those that have just changed their posture at the same position.
As described above, the image of a fish whose posture has been changed at the same position, for example, the body part may disappear and only the head and tail may remain. Even in this case, the image can be reproduced as a single fish image by the integration of the divided adjacent areas in 1) and the subsequent reduction operation in 2). Therefore, the posture change can be clearly recognized even for the fish whose posture has just been changed at the same position.

5 Fish image recognition step (S5)
Next, the fish image of the fish is recognized based on the difference image obtained as described above.
In this case, as described above, even if the fish is not moving, it can be recognized as a fish whose shape has changed if the body tilt changes or the shape changes by twisting the body. It can be distinguished from fish that have stopped working. For this reason, the presence or absence of fish activity can be determined with higher accuracy than the method of determining the presence or absence of fish activity only by whether or not the position of the fish has moved as in the prior art.
Then, a thick colored outline s (ellipse) is superimposed and displayed on the fish image of the fish recognized as moving on the screen of the monitor 3A, that is, the latest image (current image). By doing in this way, it can be judged by the presence or absence of the outline s whether the fish which has not moved completely stopped moving. The contour line s may have any shape as long as a fish image can be clearly shown. For example, an oval shape is used. Then, the color of the contour line s is changed for each fish, and when the fish moves, it is automatically tracked and always displayed with the same color of the contour line s. Note that when the fish cannot be tracked due to overlapping or shadows, another color outline s may be newly assigned. A known method may be used for the automatic tracking. For example, a method of predicting the current movement position from the previous position, speed, and direction and specifying the fish image closest to the prediction point as the same fish is used.

6 Water quality judgment step (S6)
When a fish image from which noise has been removed is obtained by performing the image recognition step, a final water quality determination step (S6) is performed.
In this step, the screen “runs” like screen a, “medaka gathers in a specific place” like screen b, “medaka moves slowly” or “stops” like screen c. Or the like is observed. By quantifying the state of the medaka f that rampages, gathers in a specific place, slows down or stops, it can be compared with the reference value, and if the reference value is also staged, Level judgments such as warnings and severe warnings can also be made.

  In the above fish image recognition method, a plurality of past images are averaged to create a background image, and by taking the difference from the target image to be monitored (latest current image), floating objects and sediment In addition, it is possible to obtain a difference image that eliminates the influence of reflection on the glass surface. And by performing image processing using this difference image, it is easy to grasp even small movements as image changes, so even if you are not moving fish, you can perform fine operations such as changing the tilt of the body Can be recognized as active. Since it is possible to monitor fine movements of fish, water quality monitoring based on fish behavior can be performed with high accuracy.

Next, a method for determining the turbidity of water that is preferably used in the water quality monitoring device of the present invention will be described.
As shown to (A) of FIG. 4, when there is no turbidity in water, the light of illumination permeate | transmits well and the brightness in the water tank 1 is substantially uniform. Further, as shown in (B), when the water becomes turbid, the light closer to the illumination is diffusely reflected and brightened, and the one far from the illumination becomes dark due to the irregular reflection. And as shown in (C), when the turbidity further enters, the tendency becomes stronger. By utilizing this property, indirect turbidity measurement from an image becomes possible.
That is, for the illumination lamp 4 that irradiates the light, the brightness ratio between the near water area and the far water area in the water tank 1 is compared, and an alarm is issued when the brightness ratio exceeds a reference value. The brightness ratio is the average value of the scattered light brightness at several points in the upper part of the monitoring area 12 in the aquarium, and the average value of the scattered light brightness at the lower several points is taken as the ratio of the upper and lower average values. It is obtained by calculating.
The reason for using a plurality of measurement points is to suppress the influence of overlapping with high brightness points such as fish shadows.
When an alarm is issued, it is possible to take measures such as changing the filter of the water channel, stopping the inflow, etc., so a false warning such as stopping the activity of the fish is sent before it becomes difficult to recognize the fish image due to strong turbidity Can be prevented.

  In the embodiment of FIG. 1, the illumination required for indirect measurement of turbidity does not require additional equipment because the lighting lamp 4 illuminating the fish can be used as it is, and it is manually or automatically depending on the measured turbidity value. Since it is possible to stop water intake, there is an advantage that it is not necessary to add complicated and expensive equipment such as a turbidimeter.

Next, a suitable fish judgment means for judging whether or not the fish in the aquarium is suitable for water quality monitoring will be described.
The first suitable fish determining means includes a first step of measuring the projected area when swimming near the center of the image captured by the camera, and whether the measured projected area is equal to or less than a lower limit value of the reference value. The second step of determining whether or not the upper limit value is exceeded.
According to the first step, the area of each inserted fish is measured when swimming near the center of the image at a set speed or higher, so that the body orientation relative to the camera and the distance from the camera are It becomes almost constant, and an accurate area can be measured. In the second step, a small fish that is smaller than the lower limit of the reference value, that is, a young fish with a small body and a poor recognition rate, or a fish that is large compared to the upper limit, that is, the body is large and the movement is slow. Because it can detect old fish, it is possible to prompt the exchange of these inappropriate fish.

Next, second suitable fish determination means will be described.
The second suitable fish determining means includes a third step of measuring acceleration during swimming of the fish and a fourth step of determining whether the sum of the absolute values of the measured accelerations is greater than or equal to a reference value. It is as follows.
On the computer 3, contour lines s of different colors are given to individual fish images, and the fish is always identified and tracked. Since the fish is identified and tracked every 0.1 second, the speed change (acceleration) in the third step can be calculated every 0.1 second.
Then, in the fourth step, the sum of absolute values of acceleration over a certain period of time is calculated, and it is determined whether or not the sum is larger than a reference value. If the result is larger than the reference value, it can be determined that the movement is violent, so that replacement can be promoted as a violent fish that drives other fish more than necessary. Thus, the water quality abnormality detection accuracy can be improved by exchanging fish that are not suitable for monitoring.

As described above, a monitor screen to be displayed on the monitor 3A can be arbitrarily adopted. Preferred monitor screens are exemplified below.
(1) Real-time monitor screen As shown in FIG. 5, a current (real-time) image is displayed every 0.1 seconds. f is a fish and s is a contour line. On the real-time monitor screen, the image used for image processing and the fish recognition result may be displayed simultaneously. Note that an alarm monitor screen and a trend monitor screen, which will be described later, may be displayed by switching the screens, or may be displayed at the same time.
(2) Alarm monitor screen Displays the number of activities and activity status of fish f in stages. For example, it is as follows.
Number of activities: Number of active fish (including tremors) Activity status: According to the conditions of detection target, detection accuracy and detection time, the optimal combination can be selected from the behavioral characteristics of the fish, and the status can be displayed in several stages.
(3) Trend monitor screen The time is displayed on the horizontal axis, and the value of each diagnostic item for the number of activities, activity state, and transparency is displayed on the vertical axis. For the display of the time axis, it is preferable to change the 1-hour display mode and the 24-hour display mode. On the screen, the yellow line is displayed as a warning line, and the red line is displayed as an alarm line. If the line exceeds or falls below this level, the green lamp displayed on the alarm monitor may be turned off or the lamp may be lit in yellow or red. .
The above is merely an example, and a monitor screen other than the above is not denied.

  The present invention is most suitable for facilities that require strict water quality management, such as water purification plants, sewage treatment facilities, drinking water manufacturers, and food manufacturers.

It is explanatory drawing of the water quality monitoring apparatus which concerns on one Embodiment of this invention. It is explanatory drawing of the image transmission method. It is explanatory drawing of an image recognition step and a water quality judgment step. It is explanatory drawing of a turbidity judgment method. It is explanatory drawing of a monitor screen.

Explanation of symbols

1 Aquarium 2 Camera 3 Computer 4 Lighting

Claims (5)

An aquarium with fish for water quality monitoring,
A camera that photographs the water tank at regular time intervals;
A computer that judges water quality based on images taken by the camera;
The computer
Image recognition means for recognizing fish images;
Water quality judging means for judging the suitability of the water quality from the recognized fish image,
The image recognition means includes
Selecting a plurality of image groups captured by the computer from the camera;
Creating a background image by averaging the luminance of each pixel in the selected image group;
Selecting a target image to be monitored and judged;
Subtracting the background image from the target image to obtain a difference image;
And a step of recognizing a fish image of an active fish based on the difference image. A water quality monitoring device equipped with an illumination lamp for illuminating the aquarium,
The computer includes turbidity determination means for water in the aquarium;
The turbidity judging means
Detecting a luminance ratio between a water area close to and far from the illumination lamp in the water tank;
The water quality monitoring apparatus according to claim 1, further comprising a step of determining whether the luminance ratio is equal to or higher than a reference value. The computer includes a suitable fish judging means for judging whether or not the fish in the aquarium is suitable for water quality monitoring;
The suitable fish judging means is
Measuring the projected area of the fish when the fish is swimming near the center of the image in which the camera is photographing the aquarium;
The water quality monitoring apparatus according to claim 1, further comprising a step of determining whether the measured projected area is equal to or less than a lower limit value or an upper limit value of the reference value. The computer includes a suitable fish judging means for judging whether or not the fish in the aquarium is suitable for water quality monitoring;
The suitable fish judging means is
Measuring acceleration during swimming of the fish;
The water quality monitoring apparatus according to claim 1, further comprising a step of determining whether a sum of absolute values of measured accelerations is equal to or greater than a reference value. A method for recognizing a fish image by photographing a water tank containing a fish for water quality monitoring with a camera at regular time intervals, and processing the obtained image with a computer.
Selecting a plurality of image groups captured by the computer from the camera;
In the selected group of images, a step of creating a background image by averaging the luminance and the like of each pixel;
Selecting a target image to be monitored and judged;
Subtracting the background image from the target image to obtain a difference image;
And recognizing a fish image of an active fish based on the difference image.