Feeding system and learning model generating method

Abstract

To improve efficiency of feeding.SOLUTION: A feeding system comprises: feeding means 2 for feeding an aquatic organism L; imaging means 3 for imaging the aquatic organism L to generate picture data; imaging control means for controlling the imaging means 3 so as to generate first picture data by imaging the aquatic organism L when an amount of the feeding means 2 feeding the aquatic organism L is a first amount and controlling the imaging means 3 so as to generate second picture data by imaging the aquatic organism L when an amount of the feeding means 2 feeding the aquatic organism L is a second amount; comparison means for comparing the first picture data generated by the imaging means 3 with the second picture data generated by the imaging means 3; and feeding control means for controlling the feeding means 2 on the basis of a processing result of the comparison means.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present disclosure relates to a feeding system and a learning model generation method.

Background Art

[0002] Conventionally, a feeding system for feeding cultured fish has been used. Patent Document 1 discloses an imaging unit that images a fish basket at a predetermined timing before, during, or after feeding. Further, Patent Document 1 discloses detecting at least one of the rate at which an organism eats food and the amount of food left uneaten by the organism by performing known image recognition processing on an imaging image during feeding.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the above conventional invention, there is room for improvement in improving the feeding efficiency.

[0005] The present disclosure aims to improve the feeding efficiency.

Means for Solving the Problems

[0006] A feeding system according to an aspect of the present invention includes a feeding means for feeding an aquatic organism, an imaging means for imaging the aquatic organism to generate image data, and when the amount of food fed by the feeding means to the aquatic organism is a first amount, controlling the imaging means to image the aquatic organism to generate first image data, and when the amount of food fed by the feeding means to the aquatic organism is a second amount, controlling the imaging means to image the aquatic organism to generate second image data; an imaging control means; a comparison means for comparing the first image data generated by the imaging means with the second image data generated by the imaging means; and a feeding control means for controlling the feeding means based on the processing result of the comparison means.

Advantages of the Invention

[0007] According to the present disclosure, the feeding efficiency can be improved.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Mode for Carrying Out the Invention

[0009] Hereinafter, with reference to FIGS. 1 to 14, the feeding system 1 according to an embodiment of the present disclosure will be described. In the following description, “based on XX” means “based at least on XX”, and includes cases where it is based on another element in addition to XX. Further, “based on XX” is not limited to the case where XX is directly used, and includes cases where it is based on something obtained by performing arithmetic operations or processing on XX. “XX” is an arbitrary element (for example, arbitrary information).

[0010] The feeding system 1 is a system related to feeding of the aquatic organism L. Examples of the aquatic organism L include fish, shrimps, crabs, etc. The aquatic organism L may be an aquatic organism L to be cultured, such as cultured fish. The aquatic organism L may be cultured in the fishpond 20.

[0011] As shown in FIGS. 1 and 2, the feeding system 1 includes a feeding means 2, an imaging means 3, a control device 4, and a learning device 5. The feeding means 2 feeds the aquatic organism L. The imaging means 3 images the aquatic organism L. The imaging means 3 generates image data. The control device 4 controls the feeding means 2 and the imaging means 3. The learning device 5 acquires information from the control device 4. The learning device 5 generates a learning model 6 from the acquired information.

[0012] As shown in FIG. 2, the feeding means 2 floats on the water in the fish basket 20. The feeding means 2 is arranged on the floating body 21. The floating body 21 is, for example, a raft. The feeding means 2 is provided with a feeding hole (not shown). The feeding means 2 discharges food from the feeding hole. The feeding hole is provided below the feeding means 2. The feeding means 2 discharges food downward (into the water) from the feeding hole through the floating body 21. The feeding means 2 can adjust the amount of food fed to the aquatic organism L (hereinafter also referred to as the feeding amount). The feeding amount is the amount of food discharged from the feeding means 2. The feeding amount may be the amount of food fed per unit time.

[0013] Note that a reference may be provided for the feeding means 2. The reference is, for example, a reference used by the control device 4 to acquire the degree of turbidity as described later. The reference may be a mark provided on the lower surface of the feeding means 2. The reference may be provided on the floating body 21 instead of the feeding means 2. The reference may be provided separately from the feeding means 2 and the floating body 21.

[0014] The imaging means 3 is arranged in the water. The imaging means 3 is located below the feeding means 2. In the illustrated example, the imaging means 3 is housed in the housing 22. The housing 22 is suspended from the floating body 21. The housing 22 is connected to the floating body 21 via a rope 23 (for example, a rope). A weight 24 is provided on the housing 22, but the weight 24 may be omitted. The relative positional relationship between the feeding means 2 and the imaging means 3 is substantially fixed. In other words, although the relative positional relationship between the feeding means 2 and the imaging means 3 may change due to, for example, tidal currents, this change has substantially no influence on various processes to be described later in the image data generated by the imaging means 3. The imaging means 3 images the aquatic organism L by imaging the feeding means 2 at a predetermined angle of view including the feeding hole.

[0015] The imaging means 3 images above the imaging means 3. The imaging means 3 images the space within the fish basket 20, which includes the feeding hole. Note that the imaging by the imaging means 3 is not limited to imaging the above space from below, and the above space may be imaged from the side (horizontally) or from above (on the sea). Also, the feeding direction or position by the feeding means 2 may be changed according to the imaging direction by the imaging means 3. For example, the feeding by the feeding means 2 may be performed from above the sea or from within the sea. Here, the feeding direction or position by the feeding means 2 may be changed independently of (without adapting to) the imaging direction by the imaging means 3. The imaging means 3 may be capable of imaging a moving image, may be capable of imaging a still image, or may be capable of imaging both a moving image and a still image.

[0016] As shown in FIG. 1, the control device 4 communicates with each of the feeding means 2, the imaging means 3, and the learning device 5 via the network NW. The network NW includes, for example, the Internet, a LAN (Local Area Network), a wireless base station, a provider device, and the like. In the example shown in FIG. 2, the imaging means 3 is wired-connected to the feeding means 2 via the information communication line 25, and the control device 4 is connected to the imaging means 3 via the feeding means 2. However, the control device 4 may be connected to the imaging means 3 without passing through the feeding means 2.

[0017] As shown in FIG. 3, the control device 4 includes a control unit 11 including a processor 91 such as a CPU (Central Processing Unit) and a memory 92 connected by a bus, and executes a feeding program. The feeding program is a program that controls the operations of the respective functional units included in the control device 4. The control device 4 functions as a device including a user interface 10, a control unit 11, and a storage unit 12 by executing the feeding program.

[0018] More specifically, the control device 4 reads out the feeding program stored in the storage unit 12 by the processor 91 and stores the read feeding program in the memory 92. By the processor 91 executing the feeding program stored in the memory 92, the control device 4 functions as a device including a user interface 10, a control unit 11, and a storage unit 12.

[0019] The user interface 10 includes a display unit 101 and an input unit 102. The display unit 101 displays various information. The display unit 101 is configured to include an output device such as a CRT (Cathode Ray Tube) display, a liquid crystal display, an organic EL (Electro-Luminescence) display, etc. The display unit 101 may be configured as an interface that connects these output devices to the control device 4.

[0020] The input unit 102 is configured to include input terminals such as a mouse, a keyboard, and a touch panel. The input unit 102 may be configured as an interface that connects these input terminals to the control device 4. The input unit 102 receives the input of various information to the control device 4. The input unit 102 functions as a designation means for designating, for example, a designated luminance to be described later to an operator. Note that the display unit 101 and the input unit 102 may be configured as an integrated touch panel.

[0021] The control unit 11 controls the operations of the respective functional units included in the control device 4. The control unit 11 controls, for example, the operation of the display unit 101.

[0022] The storage unit 12 is configured using a non-transitory computer-readable storage medium device such as a magnetic hard disk device or a semiconductor storage device. The storage unit 12 stores various information regarding the control device 4. The storage unit 12 stores, for example, a feeding program in advance. The storage unit 12 stores, for example, a feeding plan in advance. The storage unit 12 may store, for example, a learning model 6 described later.

[0023] As shown in FIG. 4, the control unit 11 includes an imaging control means 41, a discrimination means 42, a determination means 43, a comparison means 44, a calculation means 45, a feeding control means 46, and a storage control means 47.

[0024] The imaging control means 41 controls the imaging means 3. The imaging control means 41 controls the imaging means 3 so as to generate a plurality of image data at different timings. The imaging control means 41 controls the imaging means 3 so as to generate, for example, first image data and second image data. The first image data is image data when the feeding amount is the first amount. The second image data is image data when the feeding amount is the second amount. In the present embodiment, the second image data is image data at a time later than the first image data.

[0025] The imaging control means 41 controls the imaging means 3 so as to image the aquatic organism L and generate the first image data when the feeding amount is the first amount. The imaging control means 41 controls the imaging means 3 so as to image the aquatic organism L and generate the second image data when the feeding amount is the second amount.

[0026] When the imaging unit 3 is capable of imaging a moving image, the first image data and the second image data may be still images of different frames in the same moving image. When the imaging unit 3 is capable of imaging a still image, the first image data and the second image data may be still images captured at different timings. These moving images and still images may be captured at a predetermined timing and stored in the storage unit 12. In this case, the control unit 11 can acquire the image data at a past time point by reading the moving image or still image from the storage unit 12 afterwards.

[0027] Note that the first amount and the second amount may be stored in the storage unit 12, for example. The first amount and the second amount may be, for example, the feeding amount per unit time at that time. The first amount and the second amount may not be the amount of the actually fed food, but may be the amount of the food instructed by the feeding control means 46 to the feeding means 2 described later. The first amount and the second amount may be the feeding amounts in the feeding plan. The second amount may be more or less than the first amount.

[0028] Note that the imaging control means 41 may control the imaging means 3 to newly image the aquatic organism L and generate the third image data when it is determined by the discrimination means 42 described later that a predetermined period has elapsed. When the imaging unit 3 is capable of imaging a moving image, the first image data, the second image data, and the third image data may be still images of different frames in the same moving image. When the imaging unit 3 is capable of imaging a still image, the first image data, the second image data, and the third image data may be still images captured at different timings.

[0029] The discrimination means 42 discriminates the passage of time. The discrimination means 42 discriminates whether or not a predetermined period has elapsed since the imaging unit 3 generated the first image data. The processing result of the discrimination means 42 indicates the degree of oldness (newness) of the first image data. The predetermined period may be stored in the storage unit 12 in advance, for example. The predetermined period may be input from the operator through the input unit 102, for example.

[0030] The determination means 43 determines the type of the aquatic organism L (for example, fish species). The determination means 43 acquires image data and determines the type of the aquatic organism L included in the image data. The determination means 43 may determine the type of the aquatic organism L included in the image data by, for example, a known image processing method, a learning model, or the like. The determination means 43 may determine the type of the aquatic organism L from at least one of the first image data and the second image data. Instead of providing the determination means 43, for example, the type of the aquatic organism L may be stored in the storage unit 12 in advance, or may be input from the operator through the input unit 102.

[0031] The comparison means 44 compares the first image data generated by the imaging means 3 and the second image data generated by the imaging means 3. Here, the comparison by the comparison means 44 includes not only (1) directly comparing the first image data and the second image data, but also (2) comparing the first image data with other data and changing (adjusting) the second image data based on the comparison result, and (3) changing the judgment criteria based on the second image data according to the processing result of the first image data.

[0032] The comparison means 44 outputs at least one of the following <1> to <4> as a processing result. <1> "Comparison of feature amounts in each of the two image data" <2> "Image difference between the two image data" <3> "Image data with adjusted brightness" <4> "Degree of turbidity in the image data"

[0033] Note that the comparison means 44 includes an image difference acquisition unit 51, a brightness difference acquisition unit 52, a brightness adjustment unit 53, and a turbidity degree acquisition unit 54. The image difference acquisition unit 51 relates to the above <2>. The brightness difference acquisition unit 52 and the brightness adjustment unit 53 relate to the above <3>. The turbidity degree acquisition unit 54 relates to the above <4>.

[0034] <1> "Comparison of feature amounts in each of the two image data" The comparison means 44 acquires feature amounts from the first image data and the second image data respectively. The feature amount is an amount that has a correlation with at least one of, for example, the amount of uneaten food and the feeding activity (hereinafter also referred to as "feeding activity etc."). The feature amount may be, for example, the amount of food in the image data. The more the amount of food, the more the uneaten food amount can be estimated. The feature amount may be, for example, the amount of splashes in the image data. The more the amount of splashes, the higher the feeding activity can be estimated. The feature amount may be, for example, the size of the aquatic organism L in the image data. The smaller the aquatic organism L, the more the aquatic organism L is located near the sea surface and gathered around the food, and the higher the feeding activity can be estimated. The feature amount may be, for example, the degree to which the aquatic organism L gathers around the food in the image data. The higher the degree of gathering around the food, the higher the feeding activity can be estimated. Here, the degree of gathering around the food can be calculated by known image processing. The degree of gathering around the food is calculated by, for example, the amount of wave splashes, the number of aquatic organisms L located within a predetermined range based on the food, the size of each aquatic organism L, the size of the group of aquatic organisms L formed by the gathering of the aquatic organisms L, and the like.

[0035] The comparison means 44 compares the feature amount of the first image data with the feature amount of the second image data. The comparison means 44 may, for example, calculate the ratio of the feature amount of the first image data to the feature amount of the second image data, or may calculate the difference between the feature amount of the first image data and the feature amount of the second image data.

[0036] <2> "Image difference between two image data" The image difference acquisition unit 51 acquires an image difference. The image difference is the difference between the first image data generated by the imaging means 3 and the second image data generated by the imaging means 3. The acquisition of the image difference by the image difference acquisition unit 51 is realized by, for example, a known method.

[0037] <3> "Image data with brightness adjusted" The luminance difference acquisition unit 52 acquires a luminance difference, which is the difference between the luminance of the first image data generated by the imaging means 3 and a predetermined luminance. The acquisition of the luminance by the luminance difference acquisition unit 52 is realized by, for example, a known method. The luminance of each image data varies depending on, for example, the weather. For example, the luminance of image data on a sunny day is higher than that of image data on a cloudy day. Note that the luminance of each image data is the luminance of the entire image data. The predetermined luminance is, for example, the luminance of image data on a sunny day. The predetermined luminance may be stored in advance in the storage unit 12, for example. The predetermined luminance may be input from the operator through the input unit 102, for example. When the operator designates the predetermined luminance using the input unit 102 (designation means), the predetermined luminance is surely set.

[0038] The luminance adjustment unit 53 adjusts the luminance of the image data. The luminance adjustment unit 53 adjusts the image data based on the luminance difference. The luminance adjustment unit 53 adjusts the luminance of the second image data generated by the imaging means 3 based on the luminance difference (luminance difference based on the first image data) acquired by the luminance difference acquisition unit 52. The adjustment of the luminance by the luminance adjustment unit 53 is realized by, for example, a known method.

[0039] <4> "Degree of turbidity in image data" The turbidity degree acquisition unit 54 acquires the degree of turbidity of the image data. The turbidity degree acquisition unit 54 acquires, for example, the degree of turbidity of the first image data generated by the imaging means 3. Note that the turbidity degree acquisition unit 54 may acquire the degree of turbidity based on, for example, the reference appearance included in the first image data. The reference may be imaged at a fixed position regardless of the timing when the image data is imaged. Further, the turbidity degree acquisition unit 54 may acquire the degree of turbidity based on, for example, the luminance of the first image data.

[0040] Note that the comparison means 44 may change the image data to be compared according to the determination result of the determination means 43. For example, when the determination means 43 determines that the type of the aquatic organism L is yellowtail, the comparison means 44 may compare the first image data generated by the imaging means 3 and the second image data generated by the imaging means 3 as described above.

[0041] Also, the comparison means 44 may change the image data to be compared according to the discrimination result of the discrimination means 42. For example, when the discrimination means 42 does not discriminate that a predetermined period has elapsed, the comparison means 44 may compare the first image data generated by the imaging means 3 and the second image data generated by the imaging means 3. When the discrimination means 42 discriminates that a predetermined period has elapsed, the comparison means 44 may compare the second image data generated by the imaging means 3 and the third image data generated by the imaging means 3.

[0042] The calculation means 45 calculates the feeding activity and the like. For example, the calculation means 45 calculates the feeding activity and the like based on the comparison result of the comparison means 44. The calculation means 45 calculates the feeding activity and the like based on, for example, the image difference acquired by the image difference acquisition unit 51. The calculation means 45 can calculate the feeding activity and the like based on the image difference by, for example, a known method. The calculation means 45 calculates the feeding activity and the like based on, for example, the image data. The calculation means 45 can calculate the feeding activity and the like based on the image data by, for example, a known method. The calculation means 45 calculates the feeding activity and the like based on the second image data adjusted by the luminance adjustment unit 53, for example. Note that the calculation means 45 may calculate the remaining feed amount and the like based on the comparison result of the comparison means 44. For example, the calculation means 45 may calculate at least the remaining feed amount among the remaining feed amount and the feeding activity based on the comparison result of the comparison means 44. For example, the calculation means 45 may calculate at least the feeding activity among the remaining feed amount and the feeding activity based on the comparison result of the comparison means 44.

[0043] The memory control means 47 (first memory control means, second memory control means) controls the storage in the memory unit 12.

[0044] For example, the memory control means 47 causes the memory unit 12 to store the second image data adjusted by the luminance adjustment unit 53. At this time, the memory control means 47 may store the second image data adjusted by the luminance adjustment unit 53 and the feeding activity or the like calculated by the calculation means 45 in the memory unit 12 in association with each other. Hereinafter, the information associating the second image data with the feeding activity or the like is also referred to as a first data set.

[0045] For example, the memory control means 47 stores the first image data generated by the imaging means 3 and the second image data generated by the imaging means 3 in the memory unit 12 in association with each other. In addition to associating the first image data with the second image data, the memory control means 47 may associate the first image data, the second image data, and the feeding activity or the like calculated from these image data with each other. The information associating the first image data, the second image data, and the feeding activity or the like with each other is also referred to as a second data set.

[0046] Either the first data set or the second data set may be used for machine learning.

[0047] The feeding control means 46 controls the feeding means 2. Thereby, the feeding amount by the feeding means 2 is adjusted. The feeding control means 46 may control the feeding means 2 based on an input from an operator, for example. The feeding control means 46 may control the feeding means 2 based on a preset feeding plan, for example. The feeding control means 46 may control the feeding means 2 based on the processing result of the comparison means 44, for example. The feeding means 2 may control the feeding means 2 based on the processing result of the learning model 6 described later, for example.

[0048] The learning device 5 generates a learning model 6. The learning device 5 generates the learning model 6 based on the data sets (first data set, second data set) stored in the storage unit 12 of the control device 4. The hardware configuration of the learning device 5 may be the same as the hardware configuration of the control device 4.

[0049] (Feeding method) Hereinafter, an example of the feeding method by the feeding system 1 will be described. However, the feeding method is not limited to this example.

[0050] (First feeding flow) The feeding method performs, for example, a first feeding flow as shown in FIG. 5 (S10). In this first feeding flow, first, the feeding system 1 executes the feeding plan (S11). At this time, first, the feeding control means 46 reads out the feeding plan from the storage unit 12. The feeding control means 46 controls the feeding means 2 based on the feeding plan.

[0051] Next, the feeding control means 46 determines whether the plan has ended (S12). If it is determined that the plan has ended (S12: Yes), the feeding flow ends. If it is determined that the plan has not ended (S12: No), the feeding control means 46 determines whether there is a change in the feeding amount (S13). Here, the fact that there is a change in the feeding amount may be, for example, that the feeding amount at the current time is different from the feeding amount at a past time a predetermined time before the current time. The predetermined time may be, for example, stored in advance in the storage unit 12, or may be input by the operator through the input unit 102.

[0052] If there is no change in the feeding amount (S13: No), the feeding plan is continuously executed (S11). If there is a change in the feeding amount (S13: Yes), the control device 4 acquires two pieces of image data (S14). At this time, the imaging control means 41 controls the imaging means 3, and the imaging means 3 acquires the first image data and the second image data. After that, based on the acquired image data, the control device 4 considers the future feeding plan (S15). After reflecting the considered feeding plan (S16), the control device 4 continues to implement the feeding plan (S11).

[0053] (Consideration of Feeding Plan) In the consideration of the feeding plan (S15), for example, a plan consideration flow as shown in FIG. 6 is implemented. In this plan consideration flow, first, the comparison means 44 compares the first image data and the second image data (S20). By comparing the image data by the comparison means 44, the specification values required for future plan formulation are calculated. Therefore, after this comparison, the control device 4 formulates a future feeding plan based on the specification values (S30).

[0054] Here, the feeding system 1 according to the present embodiment is configured to be able to implement four methods as the comparison of image data (S20). The four methods are <1> feature amount comparison (S21), <2> image comparison (S22), <3> brightness adjustment (S23), and <4> turbidity acquisition (S24). For example, an operator can set one of these four comparison methods as the method to be implemented by the feeding system 1. Hereinafter, each comparison method and the method of formulating a plan based on the specification values obtained by the comparison method will be described in order.

[0055] <1> Feature Amount Comparison In the feature amount comparison, for example, a feature amount comparison flow as shown in FIG. 7 is implemented (S21). In this feature amount comparison flow, the comparison means 44 acquires the feature amounts of the respective image data (S21a). After that, the comparison means 44 compares the feature amounts of the respective image data (S21b). Returning to FIG. 6, the control device 4 formulates a future feeding plan based on the comparison result of the feature amounts (S31).

[0056] A specific example of this comparison and plan formulation method will be described with reference to FIGS. 8 and 9. Figures 8 and 9 are diagrams showing the relationship between the feeding plan and the feature amounts of the respective image data. Figures 8 and 9 show different examples. In any of the examples of Figures 8 and 9, two graphs ((a) and (b)) are arranged vertically. The upper graph (a) is a graph showing the relationship between time (horizontal axis) and the amount of feed (vertical axis). The lower graph (b) is a graph showing the relationship between time (horizontal axis) and the feature amount (vertical axis). In each of these graphs, the scale of the vertical axis is an example, and the scale can be changed as appropriate. The feature amount is, for example, a feature amount obtained from the image data at each time. In the illustrated example, the feature amount is a feature amount indicating the feeding activity, and the larger the feature amount, the higher the appetite of the aquatic organism L (that is, the more feeding is required). In any of the examples of Figures 8 and 9, first, on the premise of a feeding plan in which the amount of feed is 10 and after a predetermined time has elapsed, the amount of feed becomes 5. Then, when the time becomes t after the amount of feed has become 5, a review of the feeding plan is carried out. At this time, for example, the first amount of feed is 10, and the second amount of feed is 5. The first image data is image data when the amount of feed is 10, and the second image data is image data when the amount of feed is 5.

[0057] In the first example shown in Figure 8, the feature amount obtained from the first image data is 100, while the feature amount obtained from the second image data is 60. The comparison means 44 compares, for example, these ratios or differences with a preset reference value. In this case, for example, assuming that as a result of the comparison, it is determined that the feature amount (60) obtained from the second image data is sufficiently large. Then, the control device 4 formulates a plan to continue the amount of feed 5 for a predetermined time and then lower the amount of feed to 3.

[0058] On the other hand, in the second example shown in FIG. 9, the feature amount obtained from the first image data is 100, while the feature amount obtained from the second image data is 40. In this case, for example, the comparison means 44 compares these ratios or differences with a preset reference value, and as a result, determines that the feature amount (40) obtained from the second image data is small. Then, the control device 4 formulates a plan to immediately reduce the feeding amount from 5 to 3.

[0059] In this specific example, the first amount is larger than the second amount. However, the second amount may be larger than the first amount. That is, the second amount is larger than the first amount, and the feeding amount at the timing when the second image data is captured may be larger than the feeding amount at the timing when the first image data is captured. In this case, for example, the feeding means 2 increases the feeding amount from the first amount to the second amount, and the comparison means 44 compares the first image data and the second image data, so that, for example, the degree to which the aquatic organism L gathers around the bait can be calculated. Thereby, for example, the feeding activity of the aquatic organism L can be calculated with higher accuracy. For example, when the first amount is 0 and the second amount is greater than 0, the image data before feeding or during the interval between feedings (after feeding) with a feeding amount of 0 becomes the first image data, and the image data during feeding with a feeding amount exceeding 0 becomes the second image data. Also, for example, when the first amount is 0 and the second amount is greater than 0, the feeding is deliberately stopped during feeding, and after a certain period of time, the feeding is resumed (for example, a dedicated mode for calculating the feeding activity is implemented for calculating the feeding activity). In this case, the image data during the feeding stop with a feeding amount of 0 becomes the first image data, and the image data after the resumption of feeding becomes the second image data. Further, for example, when both the first amount and the second amount are greater than 0, during feeding with the second amount, the feeding amount is intentionally reduced to the first amount, and after a certain period of time, the feeding amount is returned to the second amount (for example, for calculating the feeding activity, a dedicated mode for calculating the feeding activity is implemented). In this case, the image data when the feeding amount is reduced to the first amount becomes the first image data, and the image data when the feeding amount is restored becomes the second image data.

[0060] <2> Image comparison In image comparison, for example, an image comparison flow as shown in FIG. 10 is implemented (S22). In this image comparison flow, the comparison means 44 acquires the image difference between the first image data and the second image data (S22a). Then, the comparison means 44 calculates the feeding activity and the like based on the image difference (S22b). Returning to FIG. 6, the control device 4 formulates a future feeding plan based on the feeding activity and the like (S32). For example, when the remaining food amount is large, the control device 4 formulates a plan to reduce the feeding amount, and when the remaining food amount is small, the control device 4 formulates a plan to increase the feeding amount. For example, when the feeding activity is high, the control device 4 formulates a plan to increase the feeding amount and the feeding period, and when the remaining food amount is small, the control device 4 formulates a plan to reduce the feeding amount and the feeding period.

[0061] In this case, the image difference acquisition unit 51 acquires the image difference. By using this image difference, for example, in the first image data and the second image data, the difference in the feature amount representing the remaining food amount (for example, food) and the difference in the feature amount representing the feeding activity (for example, water splashes) can be acquired. Thereby, for example, the remaining food amount, the feeding activity, and the like can be calculated with higher accuracy.

[0062] <3> Brightness adjustment In brightness adjustment, for example, a brightness adjustment flow as shown in FIG. 11 is implemented (S23). In this brightness adjustment flow, the comparison means 44 compares the brightness of the first image data with a predetermined brightness (S23a). Then, the comparison means 44 adjusts the brightness of the second image data based on the comparison result (S23b). Then, the comparison means 44 calculates the feeding activity or the like based on the adjusted second image data (S23c). Returning to FIG. 6, the control device 4 formulates a future feeding plan based on the feeding activity or the like, as in the case of image comparison (S32).

[0063] In this case, the brightness adjustment unit 53 adjusts the brightness of the second image data based on the brightness difference. The brightness difference is the difference between the brightness of the first image data and the predetermined brightness. Therefore, for example, by setting the brightness of the image data on a sunny day as the predetermined brightness, the second image data can be adjusted to an appropriate brightness regardless of the weather. As a result, the remaining food amount, the feeding activity, etc. can be calculated with higher accuracy.

[0064] <4> Turbidity acquisition In turbidity acquisition, for example, a turbidity acquisition flow as shown in FIG. 12 is executed (S24). In this turbidity acquisition flow, the turbidity degree acquisition unit 54 acquires the turbidity degree from the first image data (S24a). Then, the comparison means 44 calculates the remaining food amount based on the second image data (S24b). At this time, similar to the brightness adjustment flow, the remaining food amount may be calculated based on the second image data whose brightness has been adjusted. Returning to FIG. 6, the control device 4 formulates a future feeding plan based on the degree of turbidity and the remaining food amount (S33).

[0065] Here, a specific example of the method for formulating a plan when turbidity acquisition is assumed will be described. For example, the storage unit 12 stores in advance a predetermined threshold value related to the amount of remaining food. This predetermined threshold value is used for the determination to reduce the feeding rate or stop the feeding. If the amount of remaining food exceeds the predetermined threshold value, it is possible to determine that the food is sufficiently supplied. In this case, the feeding control means 46 reduces the feeding rate by the feeding means 2 or stops the feeding. When using such a predetermined threshold value, the higher the predetermined threshold value, the lower the feeding rate is reduced or the feeding is stopped only when the amount of remaining food (for example, the number of food items) is calculated to be large. Therefore, in this case, the possibility of reducing the feeding rate or stopping the feeding is reduced. Conversely, when the predetermined threshold value decreases, the possibility of reducing the feeding rate or stopping the feeding increases.

[0066] Furthermore, the feeding control means 46 varies the predetermined threshold value according to the degree of turbidity acquired by the turbidity degree acquisition unit 54. For example, as the degree of turbidity acquired by the turbidity degree acquisition unit 54 increases, the predetermined threshold value may be decreased. Here, when the degree of turbidity increases, the accuracy of calculating the amount of remaining food tends to decrease. Therefore, as the degree of turbidity acquired by the turbidity degree acquisition unit 54 increases, the predetermined threshold value decreases. Thus, even if the degree of turbidity increases and the accuracy of calculating the amount of remaining food decreases, excessive feeding can be suppressed. Thereby, the feeding amount can be appropriately adjusted according to the degree of turbidity. Note that, as will be described later, in the case where image processing or machine learning is assumed, when determining that an object included in an image is food if the probability that the object is food is equal to or greater than a predetermined threshold value, the predetermined threshold value may also be decreased as the degree of turbidity increases.

[0067] (Second feeding flow) The feeding method may implement a second feeding flow as shown in FIG. 13 instead of the first feeding flow as shown in FIG. 5, for example (S10A). In this second feeding flow, the method of acquiring two pieces of image data is different from that of the first feeding control flow (S14A). When acquiring two pieces of image data, first, the discrimination means 42 determines whether or not a predetermined period has elapsed (S17). When it is not determined by the discrimination means 42 that the predetermined period has elapsed (S17: No), the first image data is relatively new data generated within the predetermined period. Therefore, in this case, the comparison means 44 compares the first image data with the second image data (S18). On the other hand, when it is determined by the discrimination means 42 that the predetermined period has elapsed (S17: Yes), the comparison means 44 adopts the third image data instead of the first image data, and compares the second image data with the third image data (S19). Thereby, even if feeding is performed over a long period, the comparison means 44 can compare two pieces of image data captured at relatively close intervals. The comparison means 44 compares the first image data and the second image data. In this way, by comparing a plurality of pieces of image data captured at different timings by the comparison means 44, for example, the remaining amount of food related to the food contained in the image, the feeding activity related to the aquatic organism L contained in the image, etc. can be accurately calculated. Then, based on the processing result of such comparison means 44, the feeding control means 46 controls the feeding means 2. Therefore, the efficiency of feeding the aquatic organism L can be improved. Note that, for example, since salmon takes a long time from the start of feeding to the completion of feeding, the action and effect of such a feeding system 1 are remarkably achieved.

[0068] (Third feeding flow) The feeding method may perform, for example, a third feeding flow as shown in FIG. 14 (S40). In this third feeding flow, the determination means 43 determines whether or not the type of the aquatic organism L is yellowtail (S41). When it is determined that the fish species is yellowtail (S41: Yes), the first feeding flow is performed (S10). On the other hand, when it is not determined that the fish species is yellowtail (S41: No), the second feeding flow is performed (S10A).

[0069] The feeding time for the fish is short from the start to the end of feeding (for example, about 10 minutes). Therefore, for example, even before and after the start of feeding, there are few environmental changes, and the comparison between the image data before the start of feeding and the image data during feeding is effective from the perspective of a large change in the feeding amount. Therefore, for example, the feeding amount (=0) before the start of feeding is taken as the first amount, and the image data in this case is taken as the first image data, and the feeding amount during feeding is taken as the second amount, and the image data in this case is taken as the second image data. Then, the comparison means 44 compares the first image data and the second image data. In this way, by comparing a plurality of image data captured at different timings by the comparison means 44, for example, the remaining amount of food related to the food contained in the image and the feeding activity related to the aquatic organism L contained in the image can be accurately calculated. Based on the processing result of such a comparison means 44, the feeding control means 46 controls the feeding means 2. Therefore, the efficiency of feeding the aquatic organism L can be improved.

[0070] (Method for generating the learning model 6 using the comparison result) Here, in the first feeding flow and the second feeding flow described above, when considering the plan (S15), for example, the feeding activity and the like were calculated by image comparison (S22) and brightness adjustment (S23). When calculating the feeding activity and the like, it is also possible to utilize the learning model 6. In this case, the learning model 6 may be stored in the storage unit 12, or the learning model 6 may be stored in the learning device 5, and the control device 4 may read out the learning model 6 via the network. The learning model 6 can be generated, for example, by the following methods for generating the learning model 6 (the first generation method, the second generation method).

[0071] (The first generation method) The first generation method includes a feeding step, an imaging control step, a comparison step, an adjustment step, and a calculation step.

[0072] In the feeding step, the aquatic organism L is fed. In the imaging control step, when the amount of food fed by the feeding means 2 to the aquatic organism L is the first amount, the imaging means 3 is controlled to image the aquatic organism L and generate first image data. Also, when the amount of food fed by the feeding means 2 to the aquatic organism L is the second amount, the imaging means 3 is controlled to image the aquatic organism L and generate second image data. In the comparison step, the first image data generated by the imaging means 3 in the imaging control step is compared with the second image data generated by the imaging means 3 in the imaging control step. In the adjustment step, the second image data is adjusted based on the processing result of the comparison step. In the calculation step, based on the second image data adjusted in the adjustment step, the feeding activity and the like are calculated.

[0073] From the above-described feeding step to the calculation step, for example, it is realized by implementing a first feeding flow (S10). For example, when a feeding plan is implemented (S11), the feeding step is implemented. When image acquisition is implemented (S14), the imaging control step is implemented. When brightness adjustment (S23) in the comparison of image data (S20) is implemented, the comparison step, the adjustment step, and the calculation step are implemented. In this case, the brightness of the second image data is adjusted, but in the first generation method, elements other than the brightness in the second image data may be adjusted.

[0074] Here, in the brightness adjustment (S23), the memory control means 47 (first memory control means 47) stores, for example, the first data set. The first data set is information associating the second image data adjusted by the brightness adjustment unit 53 with the feeding activity and the like calculated by the calculation means 45.

[0075] The memory control means 47 stores the second image data in the memory unit 12. Here, the luminance of the second image data is adjusted by the luminance adjustment unit 53. Therefore, the memory unit 12 can store the image data with adjusted luminance. As a result, for example, when performing machine learning based on the image data stored in the memory unit 12, the accuracy of the generated learning model 6 can be improved. When the memory control means 47 stores the second image data in association with the feeding activity or the like calculated based on the second image data, the learning model 6 can be easily generated.

[0076] And the first generation method further includes a generation step. In the generation step, the learning model 6 is generated by machine learning using the second image data adjusted in the adjustment step as learning data and the feeding activity or the like calculated in the calculation step as teacher data. In this embodiment, the learning device 5 performs machine learning on the learning model 6 using the first data set.

[0077] In the comparison step, the first image data and the second image data are compared. Then, in the adjustment step, the second image data is adjusted based on the processing result of the comparison step. Based on the second image data adjusted in this way, the feeding activity or the like is calculated. Therefore, the calculation accuracy can be improved as compared with the case of calculating the feeding activity or the like based on mere unadjusted image data. Therefore, in the generation step, by generating the learning model 6 using the feeding activity or the like with high accuracy as teacher data in this way, a learning model 6 with high accuracy can be generated.

[0078] (Second generation method) The second generation method includes a feeding step, an imaging control step, a comparison step, and a calculation step, but does not include an adjustment step. These steps from the feeding step to the calculation step are realized, for example, by implementing the first feeding flow (S10). For example, when a feeding plan is implemented (S11), a feeding step is implemented. When image acquisition is implemented (S14), an imaging control step is implemented. When image comparison (S22) in the comparison of image data (S20) is implemented, a comparison step and a calculation step are implemented. In this case, the feeding activity or the like is calculated based on the image difference. However, in the second generation method, the feeding activity or the like may be calculated using elements other than the image difference.

[0079] In this generation method, in the image comparison (S22), the storage control means 47 (the second storage control means 47) stores, for example, the second data set. The second data set is information associating the first image data and the second image data with the feeding activity or the like calculated from these image data.

[0080] And the second generation method further includes a generation step. In the generation step, the learning model 6 is generated by machine learning using the first image data and the second image data as learning data and the feeding activity or the like calculated in the calculation step as teacher data. In the present embodiment, the learning device 5 causes the learning model 6 to perform machine learning using the second data set.

[0081] The storage control means 47 stores the first image data and the second image data in the storage unit 12 in association with each other. As a result, for example, when performing machine learning based on the image data stored in the storage unit 12, the accuracy of the generated learning model 6 can be improved. For example, in this case, by performing machine learning with the first image data and the second image data in a corresponding state, as the learning model 6, the first image data and the second image data are used as input data, and based on these input data, a learning model 6 that outputs the remaining amount of food and the feeding activity can be generated. Further, in this case, in addition to associating the first image data and the second image data, the memory control means 47 may associate the first image data, the second image data, and the feeding activity calculated from these image data. As a result, for example, a data set for machine learning can be generated in which a pair of the first image data and the second image data is learning data, and the feeding activity corresponding to this pair is teacher data. Also, as machine learning, ensemble learning using corresponding first image data and second image data can be adopted.

[0082] As described above, according to the feeding system 1 according to the present embodiment, the comparison means 44 compares the first image data and the second image data. In this way, by comparing the plurality of image data captured at different timings by the comparison means 44, for example, the remaining amount of food related to the food included in the image and the feeding activity related to the aquatic organism L included in the image can be accurately calculated. Then, based on the processing result of such a comparison means 44, the feeding control means 46 controls the feeding means 2. Therefore, the efficiency of feeding the aquatic organism L can be improved.

[0083] Note that the technical scope of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.

[0084] At least one of the image difference acquisition unit 51, the luminance difference acquisition unit 52, the luminance adjustment unit 53, the memory control means 47, the turbidity degree acquisition unit 54, the determination means 43, and the discrimination means 42 may be omitted. The learning device 5 may be omitted.

[0085] All or part of each function of the feeding system 1 may be realized using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). The program may be recorded on a computer-readable recording medium. A computer-readable recording medium is, for example, a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, or a storage device such as a hard disk incorporated in a computer system. The program may be transmitted via a telecommunication line.

[0086] In addition, within the scope not departing from the gist of the present invention, it is possible to appropriately replace the components in the above-described embodiment with well-known components, and the above-described modified examples may also be appropriately combined.

[0087] (Appended Note) <1>A feeding system according to an aspect of the present invention includes feeding means for feeding an aquatic organism, imaging means for imaging the aquatic organism to generate image data, imaging control means for controlling the imaging means to image the aquatic organism to generate first image data when the amount of food fed by the feeding means to the aquatic organism is a first amount, and controlling the imaging means to image the aquatic organism to generate second image data when the amount of food fed by the feeding means to the aquatic organism is a second amount, comparison means for comparing the first image data generated by the imaging means and the second image data generated by the imaging means, and feeding control means for controlling the feeding means based on the processing result of the comparison means.

[0088] The comparison means compares the first image data and the second image data. In this way, by comparing a plurality of image data captured at different timings by the comparison means, for example, the remaining amount of food related to the food included in the image and the feeding activity related to the aquatic organism included in the image can be accurately calculated. Then, based on the processing result of such comparison means, the feeding control means controls the feeding means. Therefore, the efficiency of feeding the aquatic organisms can be improved. Note that the comparison by the comparison means includes not only (1) directly comparing the first image data and the second image data, but also (2) comparing the first image data with other data and changing (adjusting) the second image data based on the comparison result, and (3) changing the judgment criteria based on the second image data according to the processing result of the first image data.

[0089] <2>In the feeding system according to <1> above, the comparison means may adopt a configuration including an image difference acquisition unit that acquires an image difference, which is the difference between the first image data generated by the imaging means and the second image data generated by the imaging means.

[0090] The image difference acquisition unit acquires an image difference. By using this image difference, for example, in the first image data and the second image data, differences in feature amounts representing the amount of remaining food (e.g., bait) and differences in feature amounts representing feeding activity (e.g., water splashes) can be acquired. As a result, for example, the amount of remaining food, feeding activity, etc. can be calculated with higher accuracy.

[0091] <3>In the feeding system according to <2> above, a configuration may be adopted in which the second amount is larger than the first amount.

[0092] The second amount is larger than the first amount, and the feeding amount at the timing when the second image data is captured is larger than the feeding amount at the timing when the first image data is captured. Therefore, for example, by increasing the feeding amount from the first amount to the second amount by the feeding means and comparing the first image data and the second image data by the comparison means, for example, the degree to which the aquatic organisms gather around the bait can be calculated. As a result, for example, the feeding activity of the aquatic organisms can be calculated with even higher accuracy. Here, the degree of aggregation around the bait can be calculated by known image processing. The degree of aggregation around the bait is calculated, for example, by the amount of splashes, the number of aquatic organisms located within a predetermined range based on the bait, the size of each aquatic organism, the size of the group of aquatic organisms formed by the gathering of aquatic organisms, and the like. For example, when the first amount is 0 and the second amount is greater than 0, the image data before feeding or during the interval between feedings (after feeding) with a feeding amount of 0 becomes the first image data, and the image data during feeding with a feeding amount exceeding 0 becomes the second image data. Also, for example, when the first amount is 0 and the second amount is greater than 0, the feeding is intentionally stopped during feeding, and after a certain period of time, the feeding is resumed (for example, a dedicated mode for calculating feeding activity is implemented for calculating feeding activity). In this case, the image data during the feeding stop with a feeding amount of 0 becomes the first image data, and the image data after the resumption of feeding becomes the second image data. Furthermore, for example, when both the first amount and the second amount are greater than 0, the feeding amount is intentionally reduced from the second amount to the first amount during feeding, and after a certain period of time, the feeding amount is returned to the second amount (for example, a dedicated mode for calculating feeding activity is implemented for calculating feeding activity). In this case, the image data when the feeding amount is reduced to the first amount becomes the first image data, and the image data when the feeding amount is restored becomes the second image data.

[0093] <4>In the feeding system according to <1> above, the comparison means includes a luminance difference acquisition unit that acquires a luminance difference, which is the difference between the luminance of the first image data generated by the imaging means and a predetermined luminance, and a luminance adjustment unit that adjusts the luminance of the second image data generated by the imaging means based on the luminance difference acquired by the luminance difference acquisition unit. A configuration characterized by this may be adopted.

[0094] The luminance of each image data varies depending on, for example, the weather. For example, the luminance of image data on a sunny day is higher than that of image data on a cloudy day. Note that the luminance of each image data is the luminance of the entire image data. Here, the brightness adjustment unit adjusts the brightness of the second image data based on the brightness difference. The brightness difference is the difference between the brightness of the first image data and a predetermined brightness. Therefore, for example, by setting the brightness of the image data on a sunny day as the predetermined brightness, the second image data can be adjusted to an appropriate brightness regardless of the weather. As a result, the amount of remaining food, feeding activity, etc. can be calculated with higher accuracy. In addition, when the operator designates a predetermined brightness using the designation means, the predetermined brightness is surely set.

[0095] <5>In the feeding system according to <4> above, a first storage control means for storing the second image data adjusted by the brightness adjustment unit in a storage unit may be further provided.

[0096] The first storage control means stores the second image data in the storage unit. Here, the brightness of the second image data is adjusted by the brightness adjustment unit. Therefore, the image data with adjusted brightness can be stored in the storage unit. As a result, for example, when performing machine learning based on the image data stored in the storage unit, the accuracy of the generated learning model can be improved. In addition, when the first storage control means stores the second image data in association with the amount of remaining food or feeding activity calculated based on the second image data, a learning model can be easily generated.

[0097] <6>In the feeding system according to any one of <1> to <5> above, when the amount of remaining food based on the second image data generated by the imaging means exceeds a predetermined threshold, the feeding control means reduces the feeding speed by the feeding means or stops the feeding. The comparison means includes a turbidity degree acquisition unit that acquires the degree of turbidity of the first image data generated by the imaging means, and the predetermined threshold decreases as the degree of turbidity acquired by the turbidity degree acquisition unit increases. A configuration may be adopted.

[0098] If the remaining amount of food based on the second image data generated by the imaging means exceeds a predetermined threshold, the feeding control means reduces the feeding rate by the feeding means or stops the feeding. Therefore, the higher the predetermined threshold, the feeding rate is reduced or the feeding is stopped only when it is calculated that the remaining amount of food (for example, the number of food items) is large. Therefore, in this case, the possibility of the feeding rate being reduced or the feeding being stopped is lowered. Conversely, when the predetermined threshold decreases, the possibility of the feeding rate being reduced or the feeding being stopped increases. Here, as the degree of turbidity increases, the accuracy of calculating the remaining amount of food tends to decrease. Therefore, as the degree of turbidity acquired by the turbidity degree acquisition unit increases, the predetermined threshold decreases. Therefore, even if the degree of turbidity increases and the accuracy of calculating the remaining amount of food decreases, excessive feeding can be suppressed. Thereby, the feeding amount can be appropriately adjusted according to the degree of turbidity. Note that the turbidity degree acquisition unit may acquire the degree of turbidity, for example, based on the standard appearance included in the first image data. The standard may be imaged at a fixed position regardless of the timing when the image data is imaged. Further, the turbidity degree acquisition unit may acquire the degree of turbidity, for example, based on the luminance of the first image data. Further, in the case where image processing or machine learning is assumed, when it is determined that an object included in an image is food if the probability that the object is food is equal to or more than a predetermined threshold, the predetermined threshold may be lowered as the degree of turbidity increases.

[0099] <7>In the feeding system according to any one of the above <1> to <6> aspects, a second storage control means for associating and storing the first image data generated by the imaging means and the second image data generated by the imaging means in a storage unit may be further provided.

[0100] The second storage control means associates and stores the first image data and the second image data in the storage unit. As a result, for example, when performing machine learning based on the image data stored in the storage unit, the accuracy of the generated learning model can be improved. For example, in this case, by performing machine learning with the first image data and the second image data in a corresponding state, as a learning model, the first image data and the second image data are used as input data, and based on these input data, a learning model that outputs the remaining food amount and the feeding activity can be generated. Further, in this case, in addition to associating the first image data and the second image data, the second memory control means may associate the first image data, the second image data, and at least one of the remaining food amount and the feeding activity calculated from these image data. Thereby, for example, a data set for machine learning can be generated in which a pair of the first image data and the second image data is learning data, and at least one of the remaining food amount and the feeding activity corresponding to this pair is teacher data. Also, as machine learning, ensemble learning using corresponding first image data and second image data can be adopted.

[0101] <8>The feeding system according to one aspect of the present invention includes a feeding means for feeding an aquatic organism, an imaging means for imaging the aquatic organism to generate image data, and when the amount of food fed by the feeding means to the aquatic organism is a first amount, controlling the imaging means to image the aquatic organism to generate first image data, and when the amount of food fed by the feeding means to the aquatic organism is a second amount, controlling the imaging means to image the aquatic organism to generate second image data, an imaging control means, a determination means for determining the type of the aquatic organism, a comparison means for comparing the first image data generated by the imaging means and the second image data generated by the imaging means when the determination means determines that the type is yellowtail, and a feeding control means for controlling the feeding means based on the comparison result of the comparison means.

[0102] The time from the start to the end of feeding for the fish is short (for example, about 10 minutes). Therefore, for example, even before and after the start of feeding, there are few environmental changes, and the comparison between the image data before the start of feeding and the image data during feeding is effective from the perspective of a large change in the amount of food. Thus, for example, the amount of food before the start of feeding (=0) is set as the first amount, and the image data in this case is set as the first image data, and the amount of food during feeding is set as the second amount, and the image data in this case is set as the second image data. Then, the comparison means compares the first image data and the second image data. In this way, by comparing a plurality of image data captured at different timings by the comparison means, for example, the remaining amount of food related to the food contained in the image, the feeding activity related to the aquatic organisms contained in the image, etc. can be accurately calculated. And based on the processing result of such a comparison means, the feeding control means controls the feeding means. Therefore, the efficiency of feeding the aquatic organisms can be improved.

[0103] <9>A feeding system according to an aspect of the present invention includes a feeding means for feeding an aquatic organism, an imaging means for imaging the aquatic organism to generate image data, and when the amount of food fed by the feeding means to the aquatic organism is a first amount, controlling the imaging means to image the aquatic organism to generate first image data, and when the amount of food fed by the feeding means to the aquatic organism is a second amount, an imaging control means for controlling the imaging means to image the aquatic organism to generate second image data, a determination means for determining whether or not a predetermined period has elapsed since the imaging means generated the first image data, a comparison means, and a feeding control means for controlling the feeding means based on a comparison result of the comparison means. The feeding system is characterized in that when the determination means determines that the predetermined period has elapsed, the imaging control means controls the imaging means to newly image the aquatic organism to generate third image data, and when the determination means does not determine that the predetermined period has elapsed, the comparison means compares the first image data generated by the imaging means with the second image data generated by the imaging means, and when the determination means determines that the predetermined period has elapsed, the comparison means compares the second image data generated by the imaging means with the third image data generated by the imaging means.

[0104] When the determination means does not determine that the predetermined period has elapsed, the first image data is relatively new data generated within the predetermined period. Therefore, in this case, the comparison means compares the first image data with the second image data. On the other hand, when the determination means determines that the predetermined period has elapsed, the comparison means adopts the third image data instead of the first image data and compares the second image data with the third image data. Thereby, even if feeding is carried out over a long period, the comparison means can compare two pieces of image data taken at relatively close intervals. The comparison means compares the first image data and the second image data. In this way, by comparing a plurality of pieces of image data taken at different timings by the comparison means, for example, the remaining amount of food related to the food contained in the image and the feeding activity related to the aquatic organism contained in the image can be accurately calculated. Then, based on the processing result of such comparison means, the feeding control means controls the feeding means. Therefore, the efficiency of feeding the aquatic organisms can be improved. For example, since salmon take a long time from the start to the end of feeding, the effects of such a feeding system are remarkably achieved.

[0105] <10>A learning model generation method according to an aspect of the present invention includes a feeding step of feeding an aquatic organism, and when the amount of food fed to the aquatic organism in the feeding step is a first amount, imaging the aquatic organism to generate first image data, controlling imaging means for generating image data by imaging the aquatic organism, and when the amount of food fed to the aquatic organism in the feeding step is a second amount, controlling the imaging means to image the aquatic organism to generate second image data, a comparison step of comparing the first image data generated by the imaging means in the imaging control step with the second image data generated by the imaging means in the imaging control step, an adjustment step of adjusting the second image data based on the processing result of the comparison step, a calculation step of calculating at least one of the remaining food amount and feeding activity of the aquatic organism based on the second image data adjusted in the adjustment step, and a generation step of generating a learning model by machine learning using the second image data adjusted in the adjustment step as learning data and the at least one calculated in the calculation step as teacher data.

[0106] In the comparison step, the first image data and the second image data are compared. Then, in the adjustment step, the second image data is adjusted based on the processing result of the comparison step. Based on the second image data thus adjusted, at least one of the remaining food amount and feeding activity of the aquatic organism is calculated. Therefore, the calculation accuracy can be improved as compared with the case of calculating at least one of the remaining food amount and feeding activity of the aquatic organism based on simple unadjusted image data. Therefore, in the generation step, by generating a learning model using at least one of the highly accurate residual food amount and feeding activity of aquatic organisms as teacher data, a highly accurate learning model can be generated.

Explanation of Signs

[0107] 1 Feeding system 2 Feeding means 3 Imaging means 6 Learning model 12 Memory unit 41 Imaging control means 42 Discrimination means 43 Judgment means 44 Comparison means 46 Feeding control means 47 Memory control means 47 First memory control means 47 Second memory control means 51 Image difference acquisition unit 52 Luminance difference acquisition unit 53 Luminance adjustment unit 54 Degree acquisition unit L Aquatic organism

Claims

1. Feeding means for feeding aquatic organisms, Imaging means for imaging the aquatic organisms to generate image data, Imaging control means for controlling the imaging means to image the aquatic organisms to generate first image data when the amount of food fed by the feeding means to the aquatic organisms is a first amount, and for controlling the imaging means to image the aquatic organisms to generate second image data when the amount of food fed by the feeding means to the aquatic organisms is a second amount, Comparing means for comparing the first image data generated by the imaging means with the second image data generated by the imaging means, Feeding control means for controlling the feeding means based on the processing result of the comparing means, A feeding system comprising the above.

2. The comparing means, An image difference acquisition unit for acquiring an image difference, which is the difference between the first image data generated by the imaging means and the second image data generated by the imaging means, includes, The feeding system according to claim 1, characterized in that.

3. The second amount is more than the first amount, The feeding system according to claim 2, characterized in that.

4. The comparing means, A luminance difference acquisition unit for acquiring a luminance difference, which is the difference between the luminance of the first image data generated by the imaging means and a predetermined luminance, A luminance adjustment unit for adjusting the luminance of the second image data generated by the imaging means based on the luminance difference acquired by the luminance difference acquisition unit, includes, The feeding system according to claim 1, characterized in that.

5. First storage control means for storing the second image data adjusted by the luminance adjustment unit in a storage unit, The feeding system according to claim 4, further comprising the above.

6. When the remaining food amount based on the second image data generated by the imaging means exceeds a predetermined threshold, the feeding control means reduces the feeding speed by the feeding means or stops feeding, The comparing means, A turbidity degree acquisition unit for acquiring the degree of turbidity of the first image data generated by the imaging means, includes, The predetermined threshold decreases as the degree of turbidity acquired by the turbidity degree acquisition unit increases, The feeding system according to claim 1, characterized in that.

7. Second storage control means for associating and storing the first image data generated by the imaging means and the second image data generated by the imaging means in a storage unit, The feeding system according to any one of claims 1 to 6, further comprising the above.

8. feeding means for feeding an aquatic organism; imaging means for imaging the aquatic organism to generate image data; imaging control means for controlling the imaging means to image the aquatic organism to generate first image data when the amount of food fed by the feeding means to the aquatic organism is a first amount, and for controlling the imaging means to image the aquatic organism to generate second image data when the amount of food fed by the feeding means to the aquatic organism is a second amount; determination means for determining the type of the aquatic organism; comparison means for comparing the first image data generated by the imaging means and the second image data generated by the imaging means when the type is determined to be yellowtail by the determination means; feeding control means for controlling the feeding means based on the comparison result of the comparison means; A feeding system comprising:

9. feeding means for feeding an aquatic organism; imaging means for imaging the aquatic organism to generate image data; imaging control means for controlling the imaging means to image the aquatic organism to generate first image data when the amount of food fed by the feeding means to the aquatic organism is a first amount, and for controlling the imaging means to image the aquatic organism to generate second image data when the amount of food fed by the feeding means to the aquatic organism is a second amount; discrimination means for discriminating whether or not a predetermined period has elapsed since the imaging means generated the first image data; comparison means; feeding control means for controlling the feeding means based on the comparison result of the comparison means; A feeding system comprising: When the discrimination means discriminates that the predetermined period has elapsed, the imaging control means controls the imaging means to newly image the aquatic organism to generate third image data; When the discrimination means does not discriminate that the predetermined period has elapsed, the comparison means compares the first image data generated by the imaging means and the second image data generated by the imaging means, and when the discrimination means discriminates that the predetermined period has elapsed, the comparison means compares the second image data generated by the imaging means and the third image data generated by the imaging means; A feeding system characterized by the above.

10. A feeding step of feeding an aquatic organism; When the amount of feeding the aquatic organism in the feeding step is the first amount, control an imaging means for imaging the aquatic organism to generate first image data so as to image the aquatic organism, and when the amount of feeding the aquatic organism in the feeding step is the second amount, an imaging control step of controlling the imaging means so as to image the aquatic organism to generate second image data; A comparison step of comparing the first image data generated by the imaging means in the imaging control step with the second image data generated by the imaging means in the imaging control step; An adjustment step of adjusting the second image data based on the processing result of the comparison step; A calculation step of calculating at least one of the remaining food amount and the feeding activity of the aquatic organism based on the second image data adjusted in the adjustment step; A generation step of generating a learning model by machine learning using the second image data adjusted in the adjustment step as learning data and the at least one calculated in the calculation step as teacher data; A learning model generation method characterized by comprising the above.

Images