Automatic feeding system

Abstract

To provide an automatic feeding system that can determine the amount of food that a feeder desires, such as by gradually increasing or decreasing the amount of food, whether to operate intermittently, the cycle time, and the number of repetitions, based on the amount of food calculated by some means such as AI.SOLUTION: An automatic feeding system according to the present invention includes: imaging mans that images an activity state of farmed fish; activity level determination means that determines an appetite activity level of the farmed fish; control means that determines a predetermined feeding amount on the basis of a determination result of the activity level determination unit; and feeding means that feeds the farmed fish. The control means determines the amount and speed of feeding on the basis of a predetermined amount of feeding according to the determination result, taking into account parameters arbitrarily set by the user, and operates the feeding means.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an automatic feeding system for feeding cultured fish, and particularly to an automatic feeding system capable of adjusting the feeding amount and the like.

Background Art

[0002] Conventionally, in the automatic feeding of cultured fish, there are a timer-type automatic feeder that supplies a predetermined amount of feed at a predetermined time, and a sensor-type automatic feeder that feeds by prodding a dummy feed when the fish is hungry.

[0003] However, in the timer-type automatic feeder, the feeding amount may be excessive or insufficient depending on the state of the fish and the weather. In addition, in the sensor-type automatic feeder, the feeding operation is controlled ON / OFF, and the state of the feeding operation is limited to two choices: "feeding" and "stopping". Therefore, it is impossible to feed the optimal amount corresponding to the gradually changing appetite of the fish, and there may be feed loss due to over-supply or a decrease in feeding efficiency due to insufficient supply.

[0004] Therefore, an automatic feeding system has been proposed in which artificial intelligence that has learned in advance the activity determination at the time of predation determines by analyzing the live video captured by a network camera, and controls the adjustment of the feeding amount based on the determination result (for example, Patent Document 1).

[0005] The automatic feeding system of Patent Document 1 uses artificial intelligence to utilize the knowledge, experience, intuition of fishermen engaged in the sea farming industry, sea conditions, weather data, etc. to determine the daily feeding amount and the optimal feeding time of the day, and to perform machine learning on the activity determination at the time of predation in cultured fish. Based on the result of determining the activity of the cultured fish actually being fed from the automatic feeder by the artificial intelligence, the feeding from the automatic feeder is adjusted and controlled. Thereby, it is intended to suppress the excess or deficiency of the feeding amount, perform appropriate automatic feeding, and obtain cultured fish with a uniformly sized overall shipment size according to the shipment time.

Prior Art Documents

Patent Documents

[0006] Patent Document 1 Japanese Patent No. 6739049 Summary of the Invention Problems to be Solved by the Invention

[0007] However, when making a determination by analyzing live video captured by a network camera, the movement of solids with a high appetite activity level tends to be characteristically captured. Depending on the individual differences of the fish in the fish school and the characteristics of the fish species, there is a risk that the feed will not be evenly distributed among all the cultured fish in the aquaculture farm. In addition, determination using video takes a long time to process because the data is heavy and the processing is also heavy.

[0008] In addition, many fishermen engaged in aquaculture have a desire to raise the cultured fish to their own liking and to differentiate them from the fish of others. In the above automatic feeding system, although it may be possible to raise the fish in a uniform manner, it is not possible to raise the fish with originality.

[0009] The present invention has been made in view of the above problems, and an object thereof is to solve the above problems by reflecting parameters arbitrarily set by a user in the feeding operation in addition to the appetite activity level determination information. Means for Solving the Problems

[0010] An automatic feeding system according to one aspect of the present invention includes imaging means for imaging the activity state of cultured fish, activity level determination means for determining the appetite activity level of the cultured fish, control means for determining a predetermined feeding amount based on the determination result of the activity level determination means, and feeding means for feeding the cultured fish. The control means determines a feeding amount taking into account parameters arbitrarily set by a user with respect to a predetermined feeding amount and feeding speed according to the determination result, and operates the feeding means.

[0011] According to this configuration, the activity determination means determines the appetite activity level of the cultured fish, and based on an appropriate feeding amount corresponding thereto, the feeder can set a unique feeding amount and feeding rate according to their preference, making it possible to have distinctive fish. Note that the setting of the feeding amount and feeding rate refers to the setting of the feeding amount per unit time.

[0012] Moreover, this automatic feeding system is characterized in that at least two of the parameters arbitrarily set by the user are a stepwise increase or decrease setting of the feeding amount and at least one of the presence or absence of intermittent operation, cycle time, and number of repetitions.

[0013] According to this configuration, the feeder can make fine settings, making it possible to have more distinctive fish.

[0014] In addition, in this automatic feeding system, when the control means determines a predetermined feeding amount according to the determination result of the activity determination means, it takes into account at least one of the water temperature data of the cultured water, the dissolved oxygen concentration data of the cultured water, the salinity concentration data of the cultured water, the pH value data of the cultured water, the ammonia concentration data of the cultured water, the nitrite concentration data of the cultured water, and the nitrate nitrogen concentration data of the cultured water to determine the predetermined feeding amount.

[0015] According to this configuration, when determining an appropriate feeding amount according to the determination result of the activity determination means, it is possible to determine the feeding amount including more detailed data, determine an appropriate feeding amount according to the environment, and based on this, it is possible for the feeder to set a unique feeding amount.

[0016] Furthermore, this automatic feeding system receives a signal from the imaging means, captures a reference image that serves as a reference for determining the appetite activity level of the cultured fish and a plurality of still images at a predetermined time interval within a predetermined time, and compares the still images with the reference image to determine the appetite activity level of the cultured fish.

[0017] According to this configuration, for the cultured fish, a reference image is captured, and a plurality of still images are captured at a predetermined time interval within a predetermined time. The activity level of the cultured fish towards feeding is determined based on the reference image and the plurality of still images before the cultured fish moves. Therefore, the activity level of the cultured fish towards feeding can be determined without the need for a checker to constantly monitor.

[0018] Further, this automatic feeding system is characterized in that the activity level determination means determines the appetite activity level of the cultured fish based on the reference image and the shade change of the plurality of still images.

[0019] According to this configuration, for example, the outdoor imaging environment such as the fluctuation of the light amount due to the change over time, the difference in the background around the imaging, the change in the color of the water surface due to the weather, and the presence or absence of a bird net is diverse. By determining based on the reference image and the shade change of the plurality of still images, the activity level of the cultured fish towards feeding can be accurately determined in any determination environment.

[0020] Further, this automatic feeding system includes means A for dividing the reference image and one still image into a plurality of blocks respectively, means B for calculating the difference in shade for each corresponding block of the reference image and the one still image, means C for calculating the difference in shade between the reference image and the one still image based on the calculated difference in shade for each block, and means D for determining the level of shade change based on the calculated difference in shade between the reference image and the one still image by a plurality of predetermined thresholds.

[0021] According to this configuration, by dividing the image into a plurality of blocks and determining the image based on the difference in shade for each block, the discrimination accuracy of the activity level of the cultured fish towards feeding is improved.

[0022] Further, this automatic feeding system is characterized in that means B is means for calculating the difference in the additive average value of the shade for each corresponding block of the reference image and the one still image.

[0023] According to this configuration, by calculating the average of the density fluctuations for each block calculated by means A and determining the image, the accuracy of the activity level is further improved.

[0024] Further, in this automatic feeding system, the means C uses the density difference of each block as an input value of a neural network and calculates a high-order discrimination curve for non-linear discrimination as the density difference information between the reference image and the still image of 1. The means D is characterized in that it is a means for determining the level of density change based on a plurality of predetermined thresholds for the high-order discrimination curve.

[0025] According to this configuration, by determining the image using a neural network for the density fluctuation of each block, the accuracy of determining the activity level is further improved. In addition, it is possible to determine the activity level for feeding that is not affected by the determination environment, and it is possible to make a strict and flexible determination.

[0026] Further, this automatic feeding system is characterized in that the activity level determination means has means for executing the means A to D for each of the plurality of still images at a predetermined time interval within a predetermined time for each still image.

[0027] According to this configuration, by determining the image for each of the plurality of still images, the accuracy of determining the activity level is improved as compared with the case of determining only with one still image.

[0028] Further, this automatic feeding system is characterized in that the activity level determination means determines the activity level of the cultured fish for feeding by a majority vote of each determination result of the density change level obtained by executing the means A to D for the reference image and each still image.

[0029] According to this configuration, for example, when the final determination result is the average of a plurality of determination results, there is a possibility that the final determination result deviates from reality due to a prominent determination result of 1. By calculating the final determination result by a majority vote of a plurality of determination results, the determination result of the activity level is improved.

[0030] Further, this automatic feeding system is characterized in that the majority vote of each discrimination result of each still image in the activity level determination means performs the determination of the activity level only when at least two of the discrimination results are the same, and when all of the discrimination results are different, the determination is invalidated.

[0031] According to this configuration, by invalidating when all of the plurality of determination results are different, the reliability of the determination result of the activity level is further improved.

[0032] Further, this automatic feeding system is characterized in that the neural network is learned using learning data in which a reference image and a still image corresponding to the activity level are paired for each activity level.

[0033] According to this configuration, by learning with learning data, the determination accuracy is improved each time it is used. Further, the conventional learning data was learned by associating one image data for each activity level, that is, the activity level 3 with this corresponding image data. On the other hand, in the present invention, by associating the activity level with a pair of image data such as the activity level and the corresponding image data + reference image, the determination accuracy is greatly improved compared to the conventional case. In particular, it is effective when there are fluctuations in the image according to the environment of the aquaculture farm.

[0034] Further, this automatic feeding system is characterized in that the activity level determination means performs the determination of the plurality of still images and the reference image at a predetermined time interval within a predetermined time a plurality of times at different times.

[0035] According to this configuration, by determining the still image and the reference image within a predetermined time, for example, within a time period of 5 minutes, at time intervals such as a time interval of 15 seconds, etc., the activity level during feeding can be determined, and the appropriate feeding amount can be changed at any time.

[0036] Further, this automatic feeding system is characterized in that the imaging means is disposed on the ground and the still image is an image obtained by imaging the water surface.

[0037] Here, "on the ground" means not being in water and includes areas such as the sea. According to this configuration, the water surface is imaged by the imaging means, and the activity level of the cultured fish is determined based on the image corresponding to the state of the nabla during feeding. Thereby, without installing a camera that is a foreign object in the water, the appetite activity level of the cultured fish in a natural state can be discriminated, which is preferable.

[0038] Further, this automatic feeding system is characterized in that the image obtained by the imaging means is a visible image or a near-infrared image.

[0039] According to this configuration, a system that can be utilized for various applications can be realized by signal processing using information of a visible image or a near-infrared image.

[0040] Further, instead of the imaging means, this automatic feeding system is provided with sensor irradiation means for irradiating sensor signals such as ultrasonic waves. The sensor irradiation means captures a reference reflection signal before the cultured fish moves and a plurality of reflection signals at a predetermined time interval within a predetermined time. The activity level determination means uses the reflection signal instead of the still image and uses the reference reflection signal instead of the reference image, receives the reflection signal from the sensor irradiation means, and compares the reflection signal with the reference reflection signal that serves as a reference for determining the appetite activity level of the cultured fish to determine the activity level of the cultured fish with respect to feeding.

[0041] According to this configuration, even when the imaging environment is poor, it is considered that the activity level of cultured fish with respect to feeding can be determined by the sensor irradiation means that irradiates sensor signals such as ultrasonic waves instead of the imaging means.

Advantages of the Invention

[0042] According to the automatic feeding system of the present invention, even in a situation where the imaging environment such as the location to be determined is diverse, the activity level of cultured fish with respect to feeding can be determined with high precision, and based on this, feeding according to the preferences of the feeder becomes possible.

Brief Description of the Drawings

[0043]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Modes for Carrying Out the Invention

[0044] Hereinafter, an automatic feeding system according to an embodiment of the present invention will be described with reference to the drawings, but the present invention is not limited to the following embodiments. For example, in this embodiment, an example of use in a marine cage will be described, but the present invention is not limited to marine cages and can also be applied to, for example, land cages, aquariums for land farming, and water tanks for aquariums.

[0045] <1. Configuration of the Automatic Feeding System> The automatic feeding system 1 of this embodiment determines the appetite activity level of the cultured fish in the fishpond, and based on the determination result, feeds the cultured fish with the feeding amount set by the feeder. As shown in FIG. 1, the automatic feeding system 1 mainly includes an imaging means 2, an activity level determination means 3, and a feeding means 4. In addition, the imaging means 2 and the feeding means 4 are wired-connected to the activity level determination means 3, and the tablet terminal 5 is wirelessly connected.

[0046] (Imaging Means) The imaging means 2 is an imaging device that images the activity state of the cultured fish with respect to feeding in the fishpond. Any camera can be used for the imaging means 2, for example, a UVC camera that can obtain visible images. In this embodiment, a UVC camera is used for the imaging means 2, but a digital camera, an industrial camera, a network camera, a smartphone, a tablet terminal, etc. can also be used.

[0047] The imaging means 2 is installed on the culture raft and disposed at a position where it can image the water surface where splashes occur when the cultured fish are feeding. In this embodiment, as an example, the imaging means 2 is installed at a position where it can image the water surface obliquely from above. It should be noted that as long as the imaging means 2 can be suspended from the culture raft via a cable to image the activity status of the cultured fish with respect to feeding from underwater, it is considered applicable as long as it is a position where the activity state of the cultured fish with respect to feeding can be grasped.

[0048] The imaging means 2 can adjust zoom in, zoom out, and the angle, and preferably can change the position of the imaging area. In addition, the number of the imaging means 2 is not limited to one, and a plurality of units may be used.

[0049] Then, the image data captured by the UVC camera, which is the imaging means 2, is transmitted to the activity determination means 3 for determination. Imaging and determination are repeated at regular intervals within a certain period of time to determine the level of appetite activity of the cultured fish with respect to feeding. Then, based on the principle of majority voting, the final determination of the level of appetite activity of the cultured fish with respect to feeding is made from a plurality of determination results at regular time intervals.

[0050] In this embodiment, as an example, three images are acquired at 15 - second intervals to calculate one determination result. When the same determination result appears two or more times out of three determinations using the three images, it is output as the final determination result. If all three determination results are different, the determination is considered impossible.

[0051] Then, this series of determinations (from the three determinations to the calculation of the final determination result) is repeated multiple times, for example, within 5 minutes, to continuously determine the appetite activity. Based on the determination results obtained at any time, the feeding amount is varied. Note that this time and interval are not limited to 15 seconds within 5 minutes. For example, it can also be set to make determinations every minute within 30 minutes.

[0052] (Activity determination means) Moreover, the activity determination means 3 includes a control means, and the imaging means 2, the feeding means 4, and the tablet terminal 5 are connected to it. Using this tablet terminal 5, it is possible to set and operate the activity determination means 3, as well as set and operate the imaging means 2 and the feeding means 4.

[0053] The connections to the activity determination means 3 can all be either wired or wireless. Also, the tablet terminal 5 is just an example. The operating terminal can be a PC, or it can be a terminal in a remote location using an Internet line or the like. Moreover, instead of these terminals, the activity determination means 3 may be configured to directly include a keyboard and a display.

[0054] In addition, the control means included in the activity determination means 3 is composed of a CPU, a memory, an SSD, etc. (not shown), and the determination of activity, the calculation of the feeding amount, the control of the imaging means 2 and the feeding means 4, etc. are executed by these. Further, the activity determination means 3 has means A to means E for determining the activity. These means A to means E are also executed by the control means.

[0055] Means A is a means for dividing a reference image and a single still image into a plurality of blocks respectively.

[0056] Means B is a means for calculating the difference in the weighted average value of the shading for each corresponding block between the reference image and a single still image. By inputting the difference in the weighted average value calculated by means B into a neural network, it is possible to accurately discriminate the appetite activity level of the cultured fish with respect to feeding.

[0057] Means C is a means for using the difference in the shading of each block as an input value of a neural network and calculating a high-order discrimination curve for non-linear discrimination as the difference information in the shading between the reference image and the single still image.

[0058] Means D is a means for determining the level of shading change based on a plurality of predetermined thresholds for the high-order discrimination curve.

[0059] And by using these means C and D, it is possible to accurately discriminate the appetite activity level of the cultured fish with respect to feeding. Also, by using the above means, it is possible to determine the appetite activity level of the cultured fish that is not affected by the environment of the fish farm, and it is possible to determine it precisely and flexibly.

[0060] In addition, means E is a means for executing the determination for each of the means A to D for a plurality of still images at a predetermined time interval within a predetermined time. As a result, by executing the determination for each of the plurality of still images, it is possible to accurately discriminate the appetite activity level of the cultured fish.

[0061] In addition, in this embodiment, the above determination result is calculated by being classified into activity level ranges from 0 to 5, and the result is displayed on the tablet terminal 5. Level 0 represents the state where there is no activity of the cultured fish and the activity is the lowest, and level 5 represents the state where the activity of the cultured fish is the highest. Then, according to this activity level, a preset feeding amount is calculated.

[0062] In addition, the calculation of the feeding amount from the activity level is performed by taking into account, in addition to the fish species and size of the cultured fish, water temperature data of the culture water, dissolved oxygen concentration data of the culture water, salinity concentration data of the culture water, pH value data of the culture water, ammonia concentration data of the culture water, nitrite concentration data of the culture water, and nitrate nitrogen concentration data of the culture water. Then, the optimal feeding amount for the cultured fish is calculated from these and displayed on the tablet terminal 5.

[0063] In addition, the activity determination means 3 includes a parameter setting means F for setting parameters such as the increase and decrease setting of the feeding amount, the presence or absence of intermittent operation, the cycle time setting, and the number of feeding repetitions. This setting is displayed on the tablet terminal 5 and can be arbitrarily set by the feeder. That is, the feeder can set the final desired feeding amount by taking into account the above settings in addition to the optimal feeding amount calculated by the activity determination means 3.

[0064] Note that this setting can be set each time, or these can be set in advance, and then automatic feeding can be performed for the cultured fish according to the feeding amount, intermittent operation, cycle time, number of repetitions, etc. obtained by taking these into account in addition to the calculated optimal feeding amount.

[0065] (Feeding means) The feeding means 4 in this embodiment is an automatic feeding device, which feeds the cultured fish according to a preset feeding amount, feeding time, number of feedings, etc. In addition, the feeding means 4 is connected to the activity determination means 3 and receives signals such as the desired feeding amount of the above-mentioned feeder from the activity determination means 3. Then, based on the signal, the feeding means 4 automatically feeds the cultured fish at the feeding time, feeding amount, and number of feedings.

[0066] <2. Activity determination> Next, with reference to the flowchart of FIG. 2, the process by which the activity determination means 3 determines the activity will be described.

[0067] First, when the automatic feeding system 1 is activated, an activation signal interrupts the activity determination means 3 and the system starts. Then, the imaging means 2 disposed at a position on the aquaculture raft where the water surface can be imaged images the water surface (activity level 0) in a state where no nabula is generated before the cultured fish move, and the activity determination means 3 receives and stores the image (step S1). Note that this image serves as the reference image.

[0068] Next, a predetermined time interval for imaging and determination is specified (step S2). Note that the specification of this time interval may be set in advance. FIG. 3 shows an example of the main menu screen displayed on the tablet terminal 5 connected to the activity determination means 3. In the present embodiment, as an example, the setting is to perform determination three times at intervals of Δt = 15 seconds, and an example where imaging and determination are performed three times, and the activity level is determined to be 0 three times and the final determination is displayed as the activity level 0.

[0069] Next, when feeding starts, while extracting still images at the specified time intervals (Δt = 15 seconds in the present embodiment) from the video imaged by the imaging means 2, the activity is determined using each still image. Specifically, the activity determination means 3 extracts the first still image 1 and the reference image, and performs the first motion analysis using the neural network with the first still image 1 and the reference image (step S3).

[0070] This analysis determines the appetite activity based on the shade change between the first still image 1 and the reference image. Specifically, the in-block addition average values are created for the first still image 1 and the reference image respectively after block-dividing them, the difference between these is input to the neural network, and the level of the shade change is determined.

[0071] In this neural network, as a set of two images consisting of a still image and a reference image corresponding to the activity level, multiple sets of the set of each still image and reference image corresponding to each activity level are pre-learned as learning data.

[0072] Figures 4(a) to 4(f) are diagrams showing examples of learning data from activity level 0 to activity level 5. The even-numbered (e.g., cutted-0) file names of the image data for each activity level are the reference images, and the image data with the nabla occurring in the odd-numbered (e.g., cutted-1) adjacent to the right is the still image corresponding to the activity level.

[0073] Therefore, in the case of Figure 4, it is an example of being learned with a set of three sets of reference images and still images for each activity level. Note that activity level 0 has the lowest activity level, and activity level 5 has the highest activity level.

[0074] Although not an essential configuration, in addition to the pre-learned learning data, it may be configured to accumulate and learn the actually determined data. By doing so, it is considered that further improvement of the determination system can be achieved in that environment.

[0075] Then, by inputting the data of the density change obtained from the reference image and the still image 1 into the neural network, a high-order separation curve for non-linear discrimination is obtained. This high-order separation curve is compared with the high-order separation curve according to the density change for each activity level obtained from the learning data shown in Figure 5 to determine the appetite activity level. Note that the determination in this embodiment displays the results in six levels of activity levels 0 to 5.

[0076] In this way, by performing activity determination with a neural network using the learning data learned as a pair of image data of a still image + reference image corresponding to the activity level, the determination system can be significantly improved compared to the conventional one.

[0077] That is, in the conventional learning data, the image data is one by one according to the activity level, that is, the activity level 3 is in the form of this corresponding image data. When it is determined by the threshold value for each point that the activity level and one image data are used, the possibility of misjudgment increases when there are light and dark parts protruding from a part of the image. In this embodiment, the possibility of such misjudgment can be greatly suppressed, and the determination system can be greatly improved. In particular, it is effective when the image varies greatly according to the environment of the breeding farm as in this embodiment.

[0078] After calculating the determination result of the first still image, next, a second motion analysis is performed by a neural network using the second still image 2 and the reference image Δt (15 seconds) after the still image 1 (step S4).

[0079] Furthermore, after that, a third motion analysis is performed by a neural network using the third still image 3 and the reference image 2Δt (30 seconds) after the still image 1 (step S5).

[0080] When two or more identical determination results are obtained from the three determination results, that is taken as the final determination. If all three determination results are different, the final determination is invalid and cannot be determined. Specifically, in the three - time determination of this embodiment, only when two or more identical determination results exceeding a majority are obtained among the three determination results, it is taken as the final determination result. If the determination results do not exceed a majority and are scattered, the discrimination is made invalid, and for example, "?" is displayed in the inspection result display area (step S6).

[0081] Then, the activity level of the final determination result is output to a monitor, a USB port, a file, etc. (step S7). Note that S1 to S7 are in an infinite loop, and are executed multiple times at a predetermined time interval within a predetermined time. When the predetermined time has elapsed and the feeding is finished, an end signal is sent to terminate the system.

[0082] <3. Determination of the feeding amount> When the activity level of the final determination result of one set of three times is calculated by the above-described activity determination, based on the result, the control means of the feeding amount determination means 3 determines the feeding amount. Specifically, first, from a correlation table of feeding amounts determined for each activity level, which is set in advance according to the type of cultured fish, the feeding amount corresponding to the activity level of the final determination result is selected.

[0083] Then, based on this feeding amount, by taking into account each factor of the water temperature data of the cultured water, the dissolved oxygen concentration data of the cultured water, the salinity concentration data of the cultured water, the pH value data of the cultured water, the ammonia concentration data of the cultured water, the nitrite concentration data of the cultured water, and the nitrate nitrogen concentration data of the cultured water, the optimal feeding amount is calculated. Specifically, based on the feeding amount corresponding to the above-described activity level, from the correlation table of the above-described each factor and the increase and decrease of the feeding amount, the feeding amount is increased or decreased according to the data of each factor, and the optimal feeding amount is calculated.

[0084] Note that whether each of the above-described factors can be used can be set by the feeder according to the culture environment. One of the above factors may be used, two or three factors may be used, or all factors may be used.

[0085] Then, the optimal feeding amount calculated above is displayed on the screen of the tablet terminal 5 (not shown). Also, on this screen, it is configured such that the stepwise increase and decrease of the feeding amount, the presence or absence of intermittent operation, the cycle time, and the number of repetition times can be set. Note that only one of these settings may be configured to be settable, two or three settings may be possible, or all settings may be made possible. However, it is preferable that all settings are possible so that the feeder can make detailed desired settings.

[0086] The feeder adds factors such as increasing or decreasing the feeding amount to the optimal feeding amount calculated above, determines the feeding amount and the feeding rate, i.e., the feeding amount per unit time, and sets the desired feeding in combination with the stepwise increase or decrease of the feeding amount, the presence or absence of intermittent operation, the cycle time, the number of repetitions, etc. described above. Then, the control means of the activity determination means 3 transmits the finally determined feeding amount data to the feeding means 4, and accordingly, the feeding means 4 feeds the cultured fish.

[0087] Note that this process is performed for each set of S1 - S7 of the activity determination means 3 to determine the feeding amount. The feeding amount is determined for each set of S1 - S7 that forms an infinite loop, and according to the corresponding feeding amount, the feeding means 4 feeds the cultured fish.

[0088] Also, after repeating the above process throughout the day, or for two or three days, and determining the setting of the feeding amount corresponding to each time of the day, it is possible to perform feeding based on that setting hereafter.

[0089] Furthermore, it is possible to change the stepwise increase or decrease of the feeding amount, the presence or absence of intermittent operation, the cycle time, and the number of repetitions with respect to the feeding amount corresponding to each time of the day, and the feeder can make the desired settings. In this way, the feeder can make the fish have unique characteristics.

[0090] Note that Fig. 6(a) shows an example of the optimal feeding amount and time in one - turn feeding calculated by the activity determination means 3 without setting the stepwise increase or decrease of the feeding amount, the presence or absence of intermittent operation, the cycle time, and the number of repetitions by the feeder. Fig. 6(b) shows an example of the feeding amount and time in one - turn feeding after the feeder independently sets the stepwise increase or decrease of the feeding amount, the presence or absence of intermittent operation, the cycle time, and the number of repetitions, etc.

[0091] According to the automatic feeding system 1 of the present embodiment having the above configuration, even in a situation where the imaging environment such as the place to be determined is diverse, the activity level of the cultured fish with respect to feeding can be accurately determined, and an optimal feeding amount taking into account a plurality of factors can be calculated. Furthermore, based on this, an original feeding amount according to the preference of the feeder can be set, and it becomes possible to cultivate the characteristic fish desired by the feeder.

[0092] <4. Other Embodiments> As described above, the preferred embodiments of the present invention have been described with reference to the drawings. However, it goes without saying that the present invention is not limited to the above-described embodiments, and various modifications or corrections within the scope described in the claims also belong to the technical scope of the present invention. For example, although a UVC camera is used as the imaging means in this embodiment, it is also possible to use a digital camera, an industrial camera, a network camera, a smartphone, a tablet terminal, or the like.

[0093] In addition, in the above embodiment, the activity level is determined using an image of the water surface where the state of the nabura can be seen by the camera. However, it is also conceivable to install the camera underwater and determine the activity level using an image of a fish school or the state of the fish.

[0094] In addition, in the above embodiment, the activity level is determined using a video or a still image by the camera. However, instead of the imaging means 2 that images the activity state of the cultured fish, a sensor irradiation means that irradiates a sensor signal such as ultrasonic waves is provided. The sensor irradiation means captures a reference reflection signal before the cultured fish moves and a plurality of reflection signals at a predetermined time interval within a predetermined time. The activity level determination means uses the reflection signal instead of the still image and uses the reference reflection signal instead of the reference image. It is also conceivable to apply it to a configuration in which the reflection signal from the sensor irradiation means is received and the activity level of the cultured fish with respect to feeding is determined by comparing the reflection signal with a reference reflection signal that is a reference for determining the activity level of the cultured fish with respect to feeding.

[0095] According to this, even when the imaging environment is poor, it is considered that the activity level of cultured fish with respect to feeding can be determined by sensor irradiation means that irradiates a sensor signal such as ultrasonic waves instead of the imaging means. In this case, the sensors to be used are composed of, for example, non-contact sensors such as ultrasonic sensors, electrostatic sensors, and vibrators.

Explanation of Signs

[0096] 1 Automatic feeding system 2 Imaging means 3 Activity level determination means 4 Feeding means 5 Tablet terminal

Claims

1. Imaging means for imaging the activity state of farmed fish, Activity determination means for determining the appetite activity level of farmed fish, Control means for determining a predetermined feeding amount based on the determination result of the activity determination means, Feeding means for feeding farmed fish, comprising: The control means determines the feeding amount and feeding speed taking into account the parameters arbitrarily set by the user for a predetermined feeding amount according to the determination result, and operates the feeding means, characterized in that: An automatic feeding system.

2. The parameters arbitrarily set by the user are: Stepwise feeding amount increase / decrease setting, And at least one of the presence or absence of intermittent operation, cycle time, and number of repetitions, Two or more of these, The automatic feeding system according to Claim 1.

3. The control means: When determining a predetermined feeding amount according to the determination result of the activity determination means, Taking into account at least one of the water temperature data of the culture water, dissolved oxygen concentration data of the culture water, salinity concentration data of the culture water, pH value data of the culture water, ammonia concentration data of the culture water, nitrite concentration data of the culture water, nitrate nitrogen concentration data of the culture water, to determine a predetermined feeding amount, characterized in that: The automatic feeding system according to Claim 2.

4. The activity determination means: Receives a signal from the imaging means, captures a reference image serving as a reference for determining the appetite activity level of farmed fish and a plurality of still images at a predetermined time interval within a predetermined time, Comparing the still image with the reference image to determine the appetite activity level of farmed fish, characterized in that: The automatic feeding system according to any one of Claims 1 to 3.

5. The activity determination means: In the determination of the appetite activity level of farmed fish, it is based on the reference image and the shade change of the plurality of still images. The automatic feeding system according to Claim 4.

6. The activity determination means: Means A for dividing the reference image and one still image into a plurality of blocks respectively, Means B for calculating the shade difference for each corresponding block between the reference image and the one still image, Means C for calculating the shade difference between the reference image and the one still image based on the calculated shade difference of each block, Having means D for determining the level of shade change by a plurality of predetermined thresholds for the calculated shade difference between the reference image and the one still image, The automatic feeding system according to Claim 5.

7. The means B is a means for calculating the difference between the reference image and the average value of the gradations of the corresponding blocks of the still image of 1 The automatic feeding system according to claim 6

8. The means C is a means for using the gradation difference of each block as an input value of a neural network and calculating a high-order discrimination curve of non-linear discrimination as the gradation difference information between the reference image and the still image of 1 The means D is a means for determining the level of gradation change by a plurality of predetermined thresholds for the high-order discrimination curve The automatic feeding system according to claim 6

9. The activity determination means The means A to D have means for executing each of the plurality of still images at a predetermined time interval within a predetermined time for each still image The automatic feeding system according to claim 6

10. The activity determination means For the reference image and each still image, the activity of the cultured fish with respect to feeding is determined by a majority vote of each discrimination result of the gradation change level obtained by executing the means A to D The automatic feeding system according to claim 9

11. The activity determination means The majority vote of each discrimination result of each still image The determination of the activity level is performed only when at least two of the same results are obtained for each discrimination result, When all of the discrimination results are different, the determination is invalid The automatic feeding system according to claim 10

12. The neural network is characterized in that it is learned using learning data in which a reference image and a still image corresponding to the activity level are paired for each activity level The automatic feeding system according to claim 5

13. The activity determination means The determination of the plurality of still images and the reference image at a predetermined time interval within a predetermined time is performed a plurality of times at different times The automatic feeding system according to claim 4

14. The imaging means is arranged on the ground, The still image is an image obtained by imaging the water surface The automatic feeding system according to claim 4

15. The image obtained by the imaging means is a visible image or a near-infrared image The automatic feeding system according to claim 4

16. Instead of the imaging means, a sensor irradiation means for irradiating a sensor signal such as ultrasonic waves is provided, The sensor irradiation means captures a reference reflection signal before the cultured fish moves and a plurality of reflection signals at a predetermined time interval within a predetermined time The activity determination means uses the reflection signal instead of the still image, and uses a reference reflection signal instead of the reference image, receives the reflection signal from the sensor irradiation means, and compares the reflection signal with the reference reflection signal that serves as a reference for determining the appetite activity level of the cultured fish, and discriminates the activity level of the cultured fish with respect to feeding. The automatic feeding system according to any one of claims 4 to 15.

Images