Yield prediction device, yield prediction method, computer program, learning device, learning method, and learned model

Abstract

To provide a yield prediction device, a prediction method, a program, a learning device, and a learning method for predicting a crop yield at an early stage.SOLUTION: A yield prediction device includes: a first acquisition part which acquires growth information including at least one of grass height and the number of leaves of a crop; and a second acquisition part which acquires yield information including crop yield after a specific period from now on the bases of the growth information acquired by the first acquisition part. Moreover, as the growth information, the yield prediction device acquires information further including at least one of individual leaf size, crown diameter, leaf area, and leaf projection area.SELECTED DRAWING: Figure 7

Description

[Technical field]

[0001] The present invention relates to a yield prediction technology, and more particularly to a yield prediction device, a yield prediction method, a computer program, a learning device, a learning method, and a trained model. [Background technology]

[0002] In recent years, there has been progress in the development of technology (hereinafter simply referred to as "technology") for predicting the yield of crops grown in fields or greenhouses. The yield is predicted based on image data and meteorological data (see, for example, Patent Documents 1 and 2). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-144810 A

[0004] [Patent Document 2] International Publication No. 2018 / 151196 Summary of the Invention [Problem to be solved by the invention]

[0005] The technology requires predicting crop yields at an early stage of growth.

[0006] An object of the present invention is to provide a technique capable of predicting crop yields at an early stage. [Means for solving the problem]

[0007] A yield prediction device according to one embodiment of the present invention includes a first acquisition unit that acquires growth information including at least one of a crop's plant height and the number of leaves, and a second acquisition unit that acquires yield information including the yield of the crop a specific period of time ahead based on the growth information acquired by the first acquisition unit.

[0008] A yield prediction method according to another aspect of the present invention includes obtaining growth information including at least one of the plant height and number of leaves of a crop, and obtaining yield information including the yield of the crop a specific period in the future based on the growth information.

[0009] A computer program according to another aspect of the present invention causes a computer to execute a process of acquiring growth information including at least one of the plant height and the number of leaves of a crop, and a process of predicting the yield of the crop for a specific period of time based on the growth information.

[0010] A learning device according to another aspect of the present invention includes a first acquisition unit that acquires growth information including at least one of the plant height and the number of leaves of a crop and actual yield information including at least an actual yield of the crop, a memory unit that stores a learning model in which the growth information is input information and yield information including the yield a specific period of time in the future is output information, and a learning unit that trains the learning model based on the growth information and the actual yield information previously acquired by the first acquisition unit.

[0011] A learning method according to another aspect of the present invention includes acquiring growth information including at least one of the plant height and the number of leaves of a crop and actual yield information including at least an actual yield of the crop, and training a learning model in which the growth information is input information and yield information including the yield a specific period in the future is output information, using the growth information and actual yield information acquired in the past.

[0012] A trained model according to another aspect of the present invention is generated according to the training method. Effect of the Invention

[0013] According to the present invention, a technique is provided that makes it possible to predict crop yields at an early stage. [Brief description of the drawings]

[0014] [Figure 1] FIG. 2 is a block diagram showing the configuration of the cultivation system. [Diagram 2] FIG. 2 is a schematic diagram showing the configuration of the mobile cultivation device shown in FIG. [Diagram 3] FIG. 3 is a perspective view of the cultivation bench shown in FIG. 2. [Figure 4] FIG. 4 is a cross-sectional view showing the planter shown in FIG. 3 and a crop planted in the planter. [Diagram 5] FIG. 2 is a diagram showing a detailed configuration of the prediction system shown in FIG. [Figure 6] FIG. 6 is a functional block diagram of the learning device shown in FIG. 5. [Figure 7] FIG. 6 is a functional block diagram of the prediction device shown in FIG. 5. [Figure 8] 8 is a graph illustrating the technical effect of the prediction device shown in FIG. 7. [Figure 9] FIG. 13 is a diagram showing the average value of the coefficient of determination between actual yield information and yield information for each second specific period. [Figure 10] This is a graph showing yield information from late November to July of the following year. [Figure 11] This is a graph showing actual yields under actual temperature control from early December to late March of the following year, and trends in yield information under various temperature controls. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference characters and description thereof will not be repeated.

[0016] Fig. 1 is a block diagram showing the configuration of a cultivation system 100. As shown in Fig. 1, the cultivation system 100 is a system for cultivating a crop 143 (see Fig. 4), and includes a mobile cultivation device 1, a control device 2, and a prediction system 3.

[0017] Fig. 2 is a schematic diagram showing the configuration of the mobile cultivation device 1 shown in Fig. 1. Fig. 3 is a perspective view of the cultivation bench 13 shown in Fig. 2. Fig. 4 is a cross-sectional view showing the planter 14 shown in Fig. 3 and a crop 143 planted in the planter 14.

[0018] 2 to 4, in order to facilitate understanding of the embodiment, the upward direction Z, the first direction X, and the second direction Y are each indicated by an arrow. In the embodiment, unless otherwise specified, the term "intersection" is a concept including, for example, two straight lines intersecting each other at a right angle, and, for example, two straight lines intersecting at an angle deviated from a right angle within a tolerance or fine difference range.

[0019] The upward direction Z is a vertically upward direction. The first direction X and the second direction Y intersect with each other in a plane that intersects with the upward direction Z.

[0020] As shown in Fig. 2 to Fig. 4, the mobile cultivation device 1 is installed in a horticultural facility H (see Fig. 2 in particular). The horticultural facility H is a facility for cultivating crops 143 (see Fig. 4 in particular) in places or seasons where it is difficult to grow crops. The horticultural facility H is typically a glasshouse or a vinyl greenhouse. The environmental conditions are typically temperature and / or humidity.

[0021] As shown in Fig. 2, the mobile cultivation device 1 includes a first vertical conveying device 11A, a second vertical conveying device 11B, a first horizontal conveying device 12A, a second horizontal conveying device 12B, and a plurality of cultivation benches 13. In Fig. 2, only three cultivation benches are indicated by the reference symbol "13".

[0022] The combination of the first vertical conveying device 11A, the second vertical conveying device 11B, the first horizontal conveying device 12A and the second horizontal conveying device 12B conveys a plurality of cultivation benches 13.

[0023] The first horizontal conveying device 12A and the second horizontal conveying device 12B are disposed at an interval from each other in the second direction Y. The first horizontal conveying device 12A conveys each cultivation bench 13 in one side X1 of the first direction X from the upstream part 121A to the downstream part 122A of the first horizontal conveying device 12A. The second horizontal conveying device 12B conveys each cultivation bench 13 in the other side X2 of the first direction X at a position separated in one side Y1 of the second direction Y from the second vertical conveying device 11B from the upstream part 121B to the downstream part 122B of the second horizontal conveying device 12B.

[0024] The first vertical conveying device 11A is disposed between the upstream portion 121A and the downstream portion 122B. The second vertical conveying device 11B is disposed between the downstream portion 122A and the upstream portion 121B. The first vertical conveying device 11A and the second vertical conveying device 11B are disposed to be aligned in the first direction X.

[0025] The first vertical conveying device 11A conveys each cultivation bench 13 in the other side Y2 of the second direction Y from the upstream end 111A to the downstream end 112A of the first vertical conveying device 11A. The second vertical conveying device 11B conveys each cultivation bench 13 in the one side Y1 of the second direction Y from the upstream end 111B to the downstream end 112B of the second vertical conveying device 11B.

[0026] The first vertical conveying device 11A sends each cultivation bench 13 from its downstream end 112A to the upstream part 121A of the first horizontal conveying device 12A. The first horizontal conveying device 12A sends each cultivation bench 13 from its own downstream part 122A to the upstream end 111B of the second vertical conveying device 11B. The second vertical conveying device 11B sends each cultivation bench 13 from its downstream end 112B to the upstream part 121B of the second horizontal conveying device 12B. The second horizontal conveying device 12B sends each cultivation bench 13 from its own downstream part 122B to the upstream end 111A of the first vertical conveying device 11A. Therefore, the first vertical conveying device 11A, the first horizontal conveying device 12A, the second vertical conveying device 11B and the second horizontal conveying device 12B circulate each cultivation bench 13 in the order in which they are described. In FIG. 2, the transport directions of the cultivation benches 13 are indicated by arrows A11 to A14.

[0027] As shown in FIG. 3, each growing bench 13 has a frame 131 that houses at least one planter 14.

[0028] As shown in FIG. 4, each planter 14 is a bottomed container having an upward opening 141 when accommodated on the cultivation bench 13. Each planter 14 has a shape that is elongated in the first direction X when accommodated on the cultivation bench 13 (see FIG. 3). The planter 14 accommodates soil 142 for cultivation (see FIG. 4). A crop 143 is planted in the soil 142. The crop 143 is typically strawberry, but may be another crop. A drain hole 144 is formed in the bottom of the planter 14 (see FIG. 4). Excess water, etc. is drained from the drain hole 144 to the outside of the planter 14.

[0029] The planter 14 further has a hook 146 on a periphery 145 of the opening 141 (see FIG. 4 in particular). The hook 146 is hooked onto the frame 131 (see FIG. 3 in particular), so that the planter 14 is supported so as not to shift relative to the cultivation bench 13.

[0030] Therefore, each planter 14 planted with a crop 143 is housed and supported on the cultivation bench 13 and circulated together with the cultivation bench 13 within the mobile cultivation device 1 as shown by arrows A11 to A14 (see particularly Figure 2) by the first vertical conveying device 11A, the first horizontal conveying device 12A, the second vertical conveying device 11B and the second horizontal conveying device 12B.

[0031] As shown in FIG. 2, the mobile cultivation device 1 further includes a control agent supplying device 15, a heat medium supplying device 16, a water supplying device 17, and a chemical solution supplying device 18 as various supplying devices. Since the various supplying devices can be applied with known technology, the illustration and explanation of each will be simplified. The various supplying devices operate as follows under the control of the control device 2. In particular, the control agent supplying device 15 sprays a control agent toward the soil 142 in the circulating planter 11. The heat medium supplying device 16 supplies a heat medium (hot water or cold water) for temperature adjustment toward the soil 142 on the surface portion in the circulating planter 11. The water supplying device 17 and the chemical solution supplying device 18 spray a control agent and a chemical solution, respectively, toward the soil 142 in the circulating planter 14.

[0032] The control device 2 has a memory unit 21 and a control unit 22. The memory unit 21 has a main memory device. The main memory device is, for example, a random access memory. The memory unit 21 may further have an auxiliary memory device. The auxiliary memory device includes, for example, at least one of a semiconductor memory and a hard disk drive. The memory unit 21 may include a removable medium. The memory unit 21 stores a program P1. The program P1 is a computer program for defining the operation of the mobile cultivation device 1.

[0033] The control unit 22 includes a central processing unit or a microprocessor. The control unit 22 controls the operation of the mobile cultivation device 1 by executing the program P1 in the main storage device of the storage unit 21. In detail, the control unit 22 makes each cultivation bench 13 rotate on the mobile cultivation device 1 as shown by the arrows A11 to A14 at a timing predetermined by the program P1. The control unit 22 operates various supply devices while each cultivation bench 13 is circulating.

[0034] Fig. 5 is a diagram showing a detailed configuration of the prediction system 3 shown in Fig. 1. As shown in Fig. 5, the prediction system 3 includes an imaging unit 31, a sensor unit 32, an operation unit 33, a display unit 34, a storage unit 35, and a control unit 36.

[0035] The imaging unit 31 captures images of the crops 143 in the planters 14 circulating on the mobile cultivation device 1, and generates data (hereinafter, referred to as "image data") showing the crops as a captured image by an image. In the embodiment, the imaging unit 31 has two imaging devices 311, 312. In the embodiment, each of the imaging devices 311, 312 is typically a digital video camera. However, this is not limited to this, and each of the imaging devices 311, 312 may be a digital still camera. In addition, the imaging devices 311, 312 may be a visible light camera or a camera other than visible light. In addition, the imaging devices 311, 312 may be a stereo camera capable of distance measurement, or a distance image camera capable of acquiring a distance image (also called a "distance image sensor" or "TOF camera"). Furthermore, at least one of the imaging devices 311, 312 may be a combination of two or more selected from a visible light camera, a stereo camera, and a distance image camera.

[0036] The imaging device 311 is disposed on the upper Z side of the second horizontal conveying device 12B (see also FIG. 2). Furthermore, the imaging device 311 is disposed so that the crop 143 conveyed by the second horizontal conveying device 12B is within the angle of view of the imaging device 311. Under the control of the control unit 36 ​​(i.e., the prediction device 362), the imaging device 311 captures an image of the crop 143 viewed from the upper Z direction, and periodically generates data showing the crop 143 as an image (hereinafter referred to as "first image data").

[0037] The imaging device 312 is disposed on one side Y1 of the second direction Y relative to the second horizontal conveying device 12B (see also FIG. 2). Furthermore, the imaging device 312 is disposed so that the crop 143 conveyed by the second horizontal conveying device 12B is within the angle of view of the imaging device 312. Under the control of the control unit 36, the imaging device 312 captures an image of the crop 143 viewed from one side Y1 of the second direction Y, and generates image data (hereinafter referred to as "second image data") showing an image of the crop 143. The first image data and the second image data are included in the image data.

[0038] The imaging device 311 is not limited to the above, and may be disposed on the upper side in the Z direction of the first horizontal conveying device 12A, and periodically generate first image data of the crop 143 viewed from the upper direction Z. The imaging device 312 is not limited to the above, and may be disposed on the other side Y2 in the second direction Y of the first horizontal conveying device 12A, and generate second image data of the crop 143 viewed from the other side Y2 in the second direction Y.

[0039] In the embodiment, the imaging devices 311, 312 capture images of the crops 143 transported by the first horizontal conveying device 12A or the second horizontal conveying device 12B. Therefore, the first image data and the second image data of a relatively large number of crops 143 are acquired. However, this is not limiting, and the imaging devices 311, 312 may be capable of capturing images of the crops 143 transported by the first vertical conveying device 11A or the second vertical conveying device 11B.

[0040] In the embodiment, the imaging devices 311, 312 are disposed at different positions in the upward direction Z and the second direction Y. However, it is preferable that the imaging devices 311, 312 are disposed at the same position in the first direction X. This is because, when the imaging devices 311, 312 capture images at the same time, it becomes easier to obtain first image data and second image data showing the same crop 143 viewed from different directions.

[0041] The sensor unit 32 is installed in the horticultural facility H (see FIG. 2 in particular) and generates various kinds of environmental information. The environmental information indicates at least one selected from the following (a) to (k). (a) is temperature, preferably the temperature around the crop 143 (see FIG. 4). (b) is humidity, preferably the humidity around the crop 143. (c) is illuminance, preferably the illuminance around the crop 143. (d) is carbon dioxide concentration, preferably the carbon dioxide concentration around the crop 143. (e) is saturation deficit, preferably the saturation deficit around the crop 143. (f) is dew point temperature, preferably the dew point temperature around the crop 143. (g) is wind speed, preferably the wind speed around the crop 143 (see FIG. 4). (h) is temperature of the soil 142, preferably the temperature of the soil 142 in which the crop 143 is planted. (i) is the water content, which is the water content of the soil 142. (j) is the canopy temperature of the multiple crops 143. (k) is the wetness of the leaves of the crops 143. Here, the wetness of the leaves is the wetness of the leaf surface, and is obtained by measuring the dielectric constant of the leaf surface with a known sensor.

[0042] In the embodiment, the sensor unit 32 transmits various environmental information to the control unit 36 ​​without going through a cloud server described below. Alternatively, the sensor unit 32 may upload various environmental information to a cloud server (not shown) through a network (not shown). In this case, the control unit 36 ​​downloads various environmental information from the cloud server through the network.

[0043] The operation unit 33 accepts operations by an operator and outputs information indicating the operation content (hereinafter referred to as "operation information") to the control unit 36. The operation unit 33 includes, for example, a keyboard, a touch panel, and a pointing device. The touch panel is disposed, for example, on the display surface of the display unit 34. The touch panel constitutes a graphical user interface together with the display unit 34. The operation unit 33 may be of a type that accepts user operations through communication with the outside.

[0044] The display unit 34 displays various screens under the control of the control unit 36. The display unit 34 is, for example, a liquid crystal display or an organic EL display.

[0045] The memory unit 35 has a main memory device. The main memory device is, for example, a random access memory. The memory unit 35 may further have an auxiliary memory device. The auxiliary memory device includes, for example, at least one of a semiconductor memory and a hard disk drive. The memory unit 35 may include a removable medium. The memory unit 35 stores programs P2 and P3 and a learning model M1. Depending on the situation, the learning model M1 may be described as a learned model M1.

[0046] The program P2 is a computer program that specifies the learning process of the learning model M1 by the control unit 36. In detail, the program P2 is a computer program for executing a machine learning algorithm for finding certain rules from a plurality of pieces of learning data and generating a learned model M1 that expresses the found rules.

[0047] The program P3 is a computer program that prescribes the prediction process performed by the control unit 36 ​​using the learning model M1.

[0048] In the learning model M1, growth information described below is input information, and yield information described below is output information.

[0049] The control unit 36 ​​includes a central processing unit or a microprocessor. The control unit 36 ​​executes the program P2 to function as a learning device 361 together with the storage unit 35. The control unit 36 ​​executes the program P2 to perform machine learning on a plurality of pieces of learning data, and adjust parameters of the inference program. As a result, a trained model M1 is generated.

[0050] The machine learning algorithm is not particularly limited as long as it is supervised learning, and may be, for example, a regression analysis, a decision tree, a nearest neighbor method, a naive Bayes classifier, a support vector machine, or a neural network. Therefore, the algorithm of the trained model M1 includes a regression analysis, a decision tree, a nearest neighbor method, a naive Bayes classifier, a support vector machine, or a neural network. In the machine learning for generating the trained model M1, the backpropagation method may be used.

[0051] For example, the neural network includes an input layer, one or more intermediate layers, and an output layer. Specifically, the neural network is a deep neural network (DNN), a recurrent neural network (RNN), or a convolutional neural network (CNN), and performs deep learning. For example, the deep neural network includes an input layer, multiple intermediate layers, and an output layer.

[0052] Moreover, the control unit 36 ​​executes the program P3 to function as a prediction device 362 together with the storage unit 35. Therefore, although not shown in FIG.

[0053] Fig. 6 is a functional block diagram of the learning device 361 shown in Fig. 5. Hereinafter, the learning device 361 and a learning method according to an embodiment of the present invention will be described with reference to Fig. 6. In addition to the storage unit 35, the learning device 361 includes a first acquisition unit 3611 and a second acquisition unit 3612 as functional blocks.

[0054] The first acquisition unit 3611 acquires growth information including at least one of the plant height and the number of leaves of the crop 143, and other information indicating at least the actual yield of the crop 143. The other information is an example of the "actual yield information" of the present invention. The other information may include time information indicating the number of days or weeks required to cultivate the crop 143.

[0055] In detail, the first acquisition unit 3611 acquires growth information including at least one of plant height and number of leaves for each crop 143 on the mobile cultivation apparatus 1. More specifically, the first acquisition unit 3611 acquires image data of the crop 143 on the mobile cultivation apparatus 1, and acquires the growth information based on the acquired image data. Since the growth information is based on image data, it is easy to obtain the growth information. Note that the growth information may further indicate at least one of the size of individual leaves of the crop 143, the crown diameter, the leaf area, and the leaf projected area.

[0056] More specifically, the first acquisition unit 3611 acquires first image data for each crop 143 from the imaging device 311 and acquires second image data for each crop 143 from the imaging device 312 while each cultivation bench 13 makes one revolution on the mobile cultivation device 1. The first acquisition unit 3611 performs image processing on at least one of the first image data and the second image data to specify the plant height and the number of leaves for each crop 143. Note that the number of leaves and the plant height are feature amounts of the crop 143 and are concepts different from images. Typically, the first acquisition unit 3611 acquires growth information including the plant height and the number of leaves for each crop 143 a specific number of times during a specific period from after planting of the crop 143 to the end of cultivation. Not limited to this, the first acquisition unit 3611 may acquire growth information a specific number of times during a specific period from after planting of the crop 143 to harvest. The specific period is, for example, one day. Also, the specific number of times is, for example, once. Furthermore, the planting time may be once a year, or may be multiple times a year. The first acquisition unit 3611 stores the acquired growth information in the storage unit 35 in a state linked to identification information and date information. The identification information is information for uniquely identifying the crop 143 that is the subject of the growth information. Furthermore, the date information is information indicating the date on which the growth information linked to the date information was acquired.

[0057] The size of individual leaves, the crown diameter, the leaf area, and the leaf projected area of ​​the crop 143 are obtained from at least one of the first image data and the second image data.

[0058] The plant height, the number of leaves, the size of individual leaves, the crown diameter, the leaf area, and the leaf projection area may be specified not only for each crop 143, but also for each planter 14 or each cultivation bench 13. For example, at least one of the first image data that are continuous on the time axis and the second image data that are continuous on the time axis may be synthesized by a stitching process. By using such synthesized image data, the first acquisition unit 3611 can specify each feature amount not only for each crop 143, but also for each planter 14 or each cultivation bench 13.

[0059] The first acquisition unit 3611 specifies the feature amount by image processing, but may specify the feature amount by three-dimensional data processing further using depth information restored from the first image data or the second image data.

[0060] Furthermore, the first acquisition unit 3611 does not specify the feature amount for each crop 143 using at least one of the first image data and the second image data, but obtains the feature amount for each micro-region obtained by dividing one crop 143 into a plurality of regions using at least one of the first image data and the second image data. The first acquisition unit 3611 may obtain the feature amount for each crop 143, each planter 14, or each cultivation bench 13 based on the feature amount for each micro-region.

[0061] Image processing or three-dimensional data processing may be performed not only for each crop 143, each planter 14, and each cultivation bench 13, but also for each specific time.

[0062] In addition, it is preferable for the first acquisition unit 3611 to acquire environmental information from the sensor unit 32 at a predetermined time interval in order to improve the prediction accuracy of the yield. The first acquisition unit 3611 stores the acquired environmental information in the storage unit 35 in a state linked to date information of the target environmental information. Here, the first acquisition unit 3611 may link the average value of temperature acquired on the same day to the date information as the environmental information. Similarly, the first acquisition unit 3611 may link the average value of humidity acquired on the same day or the average value of illuminance, etc. acquired on the same day to the date information as the environmental information.

[0063] Further, the first acquisition unit 3611 acquires other information indicating at least the actual yield of the crop 143 for each crop 143 on the mobile cultivation device 1. Alternatively, the first acquisition unit 3611 may acquire other information indicating at least the actual yield of the crop 143 for each specific limited area in the mobile cultivation device 1. The limited area is, for example, for each planter 14 or each cultivation bench 13. In detail, the worker operates the operation unit 33 of the learning device 361 to input other information indicating at least the actual yield of the crop 143 after the harvest of the crop 143. Alternatively, the worker may input time information. When inputting, the worker operates the operation unit 33 to specify the identification information of the harvested crop 143. As a result, the operation unit 33 transmits the other information and the identification information to the first acquisition unit 3611 in a state in which the other information and the identification information are linked to each other. The first acquisition unit 3611 stores the acquired other information and the identification information in the storage unit 35 in a state in which the acquired other information and the identification information are linked to date information.

[0064] As a result of the processing by the first acquisition unit 3611, the growth information, environmental information, and other information for each of the crops 143 are accumulated in the storage unit 35 for each date.

[0065] The second acquisition unit 3612 reads out, at a post-harvest timing designated by the operator, the growth information, the environmental information, and other information within a range from the harvest date to a first specific period from the storage unit 35 for each piece of specific information. The first specific period is, for example, five months.

[0066] When predicting yield, the history of the last five weeks at a specific point in time is called up, and feature values ​​are calculated using the history information of the last five weeks, and are trained by the learning model M1. Feature values ​​refer to arithmetic values ​​including the average value for each parameter (the above-mentioned growth information, environmental information, and other information) in each history period, the sum of the average value and twice the standard deviation, the difference between the average value and twice the standard deviation, or the calculated regression coefficient.

[0067] As described above, the memory unit 35 stores the learning model M1. More specifically, the learning model M1 has input information that is the feature amount of the growth information, and output information that includes the yield in the second specific period. Here, the yield in the second specific period is also a predicted value of the yield. The learning model M1 may also have additional input information that is a feature amount calculated from the environmental information and other information. The second specific period is an example of the "specific period" of the present invention, and is a maximum of 8 weeks. The second specific period may also be longer than 8 weeks. A specific second specific period may be input to the prediction device 362 by an operator by operating the operation unit 33 of the prediction device 362. A specific second specific period may also be recorded in advance in the program P3.

[0068] The second acquisition unit 3612 trains the learning model M1 based on the read growth information and other information, i.e., training data that is the growth information and other information previously acquired by the first acquisition unit. Environmental information may be added to the training data. By generating the training data according to the above learning method, a trained model M1 capable of predicting the yield for the second specific period ahead is provided. Furthermore, by inputting time information as other information, the prediction accuracy of the yield can be further improved.

[0069] The second acquiring unit 3612 may execute a data cleansing process on the growth information and other information before learning the learning model M1.

[0070] Fig. 7 is a functional block diagram of the prediction device 362 shown in Fig. 5. Hereinafter, the prediction device 362, a prediction method, and a computer program according to an embodiment of the present invention will be described with reference to Fig. 7. The prediction device 362 includes, in addition to the storage unit 35, a first acquisition unit 3621, a second acquisition unit 3622, and a display control unit 3623 as functional blocks.

[0071] The first acquisition unit 3621 acquires growth information including at least one of the plant height and the number of leaves of the crop 143.

[0072] The first acquisition unit 3621 may acquire the growth information in a manner similar to that acquired by the above-described first acquisition unit 3611. In particular, the first acquisition unit 3621 may acquire the growth information based on image data of the crop 143 on the mobile cultivation device 1, similar to the first acquisition unit 3611. The growth information can be easily obtained based on the image data. Note that each of the first acquisition units 3611 and 3621 may acquire the actual measured plant height and number of leaves through the operation of the operation unit 33 by the operator.

[0073] The first acquisition unit 3621 may acquire the environmental information in a manner similar to that used by the first acquisition unit 3611 described above.

[0074] The second acquisition unit 3622 acquires yield information including the yield of the crop 143 in the second specific period ahead based on the growth information acquired by the first acquisition unit 3621. This makes it possible to predict the yield of the crop 143. When the first acquisition unit 3621 acquires environmental information, the second acquisition unit 3622 may acquire yield information further based on the environmental information acquired by the first acquisition unit 3621.

[0075] In detail, the second acquisition unit 3622 provides the growth information acquired by the first acquisition unit 3621 as input information to the trained model M1 stored in the storage unit 35. Environmental information may be added as the input information. As a result, the second acquisition unit 3622 acquires yield information as output information from the trained model M1. As a result, the yield information of the crop 143 can be obtained more quickly and accurately than in the background art. In addition, by providing the growth information to the trained model M1 to obtain the yield information, the crop yield can be predicted at an earlier stage than in the background art, as will be described in detail later.

[0076] The second acquiring unit 3622 may execute a data cleansing process on the growth information and other information before learning the learning model M1.

[0077] The display control unit 3623 causes the display unit 34 to display the yield information acquired by the second acquisition unit 3622. The worker recognizes the yield information as a prediction result of the prediction device 362 through the display unit 34.

[0078] FIG. 8 is a graph showing the technical effect of the prediction device 362 shown in FIG. 7. The inventor trained the learning model M1 in the learning device 361, and confirmed yield information for each of the multiple crops 143 for eight weeks ahead, which is a second specific period ahead from a certain point in time, using the prediction device 362. The inventor also confirmed the actual yield for each of the same multiple crops 143 for eight weeks ahead from the same point in time. The yield information and actual yield are in grams. The inventor further plotted the actual yield and yield information for each of the same crops 143 on a graph with the actual yield on the horizontal axis and the yield information on the vertical axis. As a result, a large coefficient of determination R of 0.877 was found between the actual yield and the yield information. 2 It was therefore confirmed that the learning device 361 and the prediction device 362 can predict the yield of a crop as early as eight weeks in advance based on growth information including at least one of the plant height and the number of leaves of the crop.

[0079] FIG. 9 is a diagram showing the average value of the coefficient of determination between the actual yield information and the yield information for each second specific period. As shown in FIG. 9, the coefficient of determination is about 0.89 in the case of 5 weeks ahead and 6 weeks ahead, about 0.88 in the case of 7 weeks ahead and 8 weeks ahead, about 0.86 in the case of 9 weeks ahead, about 0.84 in the case of 10 weeks ahead, about 0.82 in the case of 11 weeks ahead, and about 0.83 in the case of 12 weeks ahead. According to FIG. 9, the prediction accuracy is lower in the case of 9 weeks to 12 weeks ahead than in the case of 5 weeks to 8 weeks ahead. Thus, from the results of FIG. 9, it was confirmed that the crop yield can be predicted with high accuracy at an early stage of 8 weeks in advance.

[0080] Fig. 10 is a graph showing yield information from late November to July of the following year. According to the example of the graph in Fig. 10, yield information including a larger yield compared to other months (see box W11) was obtained from the prediction device 362 around March, and yield information including a relatively smaller yield (see box W12) was obtained from the prediction device 362 during January and from late April to early May. By referring to such a graph, a harvesting work plan (such as the number of employed personnel) can be appropriately formulated.

[0081] FIG. 11 is a graph showing the actual yield under actual temperature control from early December to late March of the following year, and yield information under various temperature controls. In FIG. 11, the actual yield under actual temperature control is shown by a solid line L01. Also, yield information at the same temperature as the actual temperature control is shown by a dashed line L02. Yield information under temperature control at -0.5°C relative to the actual temperature control is shown by a dashed line L03. Yield information under temperature control at -1.0°C relative to the actual temperature control is shown by a dashed line L04. Yield information under temperature control at -1.5°C relative to the actual temperature control is shown by a dashed line L05. Yield information under temperature control at -2.0°C relative to the actual temperature control is shown by a dashed line L06.

[0082] From the graph in FIG. 11, the ratio of the yield information to the actual yield at the same temperature as the actual temperature control is approximately 100% in percentage. The ratio of the yield information to the actual yield under temperature control of -0.5°C relative to the actual temperature control is approximately 94% in percentage. The ratio of the yield information to the actual yield under temperature control of -1.0°C relative to the actual temperature control is approximately 93% in percentage. The ratio of the yield information to the actual yield under temperature control of -1.5°C relative to the actual temperature control is approximately 92% in percentage. The ratio of the yield information to the actual yield under temperature control of -2.0°C relative to the actual temperature control is approximately 91% in percentage. Therefore, in order to maximize the overall profit, a combination of temperature control in the horticultural facility H and the utility costs required for the temperature control can be confirmed.

[0083] The embodiments of the present disclosure have been described above with reference to the drawings. However, the present disclosure is not limited to the above embodiments, and can be implemented in various aspects without departing from the spirit of the present disclosure. In addition, the components disclosed in the above embodiments can be modified as appropriate. For example, a certain component among all the components shown in an embodiment may be added to a component of another embodiment, or some components among all the components shown in an embodiment may be deleted from the embodiment.

[0084] In addition, the drawings are mainly schematic illustrations of each component to facilitate understanding of the present disclosure, and the thickness, length, number, spacing, etc. of each illustrated component may differ from the actual ones due to the convenience of creating the drawings. In addition, the configurations of each component shown in the above embodiment are merely examples and are not particularly limited, and it goes without saying that various modifications are possible within a range that does not substantially deviate from the effects of the present disclosure.

[0085] In the embodiment, the learning device 361 and the prediction device 362 each execute processing for the crop 143 in the planter 14. However, this is not limited to the above, and each processing of the learning device 361 and the prediction device 362 may be executed for the crop planted in soil (i.e., the surface of the land). Here, the crop may be planted in soil either inside the horticultural facility H or outside the horticultural facility H.

[0086] In the embodiment, the crop 143 can be rotated in the directions of the arrows A11 to A14 by the mobile cultivation apparatus 1. The imaging unit 31 outputs image data showing the rotating crop 143 by an image. However, the present invention is not limited to this, and the crop 143 may be planted in soil and immovable. In this case, the imaging unit 31 is mounted on a mobile body that can move in the soil or in the air above the soil. In detail, the imaging unit 31 is mounted on a traveling device that travels on the soil, or on an unmanned aerial vehicle (a so-called drone) or a manned aerial vehicle that flies in the air above the soil. Alternatively, the imaging unit 31 may move along a rail extending above the mobile cultivation apparatus 1 in the horticultural facility H (see FIG. 2).

[0087] Moreover, the prediction system 3 is installed inside the horticultural facility H. However, this is not limiting, and the storage unit 35 and the control unit 36 ​​of the prediction system 3 may be provided in a cloud server.

[0088] This application discloses the following supplementary notes, which are not intended to limit the present invention.

[0089] (Appendix 1) A first acquisition unit that acquires growth information including at least one of a plant height and a number of leaves of a crop; a second acquisition unit that acquires yield information including a yield of the crop in a specific period ahead based on the growth information acquired by the first acquisition unit; A yield prediction device comprising:

[0090] (Appendix 2) The yield prediction device described in Appendix 1, wherein the growth information further includes at least one of individual leaf size, crown diameter, leaf area, and leaf projected area.

[0091] (Appendix 3) The first acquisition unit further acquires environmental information, the environmental information includes at least one of temperature, humidity, illuminance, carbon dioxide concentration, vapor pressure deficit, dew point temperature, wind speed, wetness of the leaves of the crop, soil temperature, water content of the soil, and canopy temperature of the crop; 3. The yield prediction device according to claim 1, wherein the second acquisition unit acquires the yield information further based on the environmental information acquired by the first acquisition unit.

[0092] (Appendix 4) A yield prediction device as described in any one of claims 1 to 3, wherein a first acquisition unit acquires image data of the crop and acquires the growth information based on the image data.

[0093] (Appendix 5) The first acquisition unit further acquires actual yield information including at least an actual yield of the crop, A storage unit that stores a learning model in which the growth information is input information and the yield information is output information; a learning unit that learns the learning model based on the growth information and the other information previously acquired by the first acquisition unit; Further comprising: The second acquisition unit acquires the yield information as the output information by inputting the growth information acquired by the first acquisition unit into the learning model as the input information.

[0094] (Appendix 6) Acquiring growth information including at least one of a plant height and a number of leaves of a crop; acquiring yield information including a yield of the crop for a specific period ahead based on the growth information; A method for predicting yield, including:

[0095] (Appendix 7) On the computer, A process of acquiring growth information including at least one of a plant height and a number of leaves of a crop; A process of predicting a yield of the crop for a specific period of time based on the growth information; A computer program that executes the following:

[0096] (Appendix 8) A first acquisition unit that acquires growth information including at least one of a plant height and a number of leaves of a crop, and actual yield information including at least an actual yield of the crop; A storage unit that stores a learning model in which the growth information is input information and yield information including the yield in a specific time period ahead is output information; a learning unit that learns the learning model based on the growth information and the actual yield information previously acquired by the first acquisition unit; A learning device comprising:

[0097] (Appendix 9) Acquiring growth information including at least one of a plant height and a number of leaves of a crop and actual yield information including at least an actual yield of the crop; A learning model in which the growth information is input information and yield information including the yield in a specific period ahead is output information is trained using the growth information and actual yield information acquired in the past. Including, how to learn.

[0098] (Appendix 10) A trained model generated according to the training method described in Appendix 9. [Industrial Applicability]

[0099] The present invention relates to a yield prediction device, a yield prediction method, a computer program, a learning device, a learning method, and a trained model, and has industrial applicability. [Explanation of symbols]

[0100] 100: Cultivation system 1: Mobile cultivation device 2: Control device 3: Prediction system 31: Imaging unit 311, 312: Imaging device 32: Sensor unit 33:Operation unit 34: Display section 35: Storage section 36: Control section 361: Learning device 3611 :1st acquisition part 3612: Learning Department 362: Prediction device 3621 :1st acquisition part 3622:Second acquisition part 3623: Display control section M1: Learning model (pre-trained model) P1~P3: Program

Claims

1. A first acquisition unit that acquires growth information including at least one of a plant height and a number of leaves of a crop; a second acquisition unit that acquires yield information including a yield of the crop in a specific period ahead based on the growth information acquired by the first acquisition unit; A yield prediction device comprising:

2. The yield prediction device according to claim 1 , wherein the growth information further includes at least one of individual leaf size, crown diameter, leaf area, and leaf projected area.

3. The first acquisition unit further acquires environmental information, the environmental information includes at least one of temperature, humidity, illuminance, carbon dioxide concentration, vapor pressure deficit, dew point temperature, wind speed, wetness of leaves of the crop, soil temperature, water content of the soil, and canopy temperature of the crop; The yield prediction device according to claim 1 , wherein the second acquisition unit acquires the yield information further based on the environmental information acquired by the first acquisition unit.

4. The yield prediction device according to claim 1 , wherein the first acquisition unit acquires image data of the crop, and acquires the growth information based on the image data.

5. The first acquisition unit further acquires actual yield information including at least an actual yield of the crop, A storage unit that stores a learning model in which the growth information is input information and the yield information is output information; a learning unit that learns the learning model based on the growth information and the actual yield information previously acquired by the first acquisition unit; Further comprising: The yield prediction device of claim 1 or claim 2, wherein the second acquisition unit acquires the yield information as the output information by providing the growth information acquired by the first acquisition unit to the learning model as the input information.

6. Acquiring growth information including at least one of a plant height and a number of leaves of a crop; acquiring yield information including a yield of the crop for a specific period ahead based on the growth information; A method for predicting yield, including:

7. On the computer, A process of acquiring growth information including at least one of a plant height and a number of leaves of a crop; A process of predicting a yield of the crop for a specific period of time based on the growth information; A computer program that executes the following:

8. A first acquisition unit that acquires growth information including at least one of a plant height and a number of leaves of a crop, and actual yield information including at least an actual yield of the crop; A storage unit that stores a learning model in which the growth information is input information and yield information including the yield in a specific time period ahead is output information; a learning unit that learns the learning model based on the growth information and the actual yield information previously acquired by the first acquisition unit; A learning device comprising:

9. Acquiring growth information including at least one of a plant height and a number of leaves of a crop and actual yield information including at least an actual yield of the crop; A learning model in which the growth information is input information and yield information including the yield in a specific period ahead is output information is trained using the growth information and actual yield information acquired in the past. Including, how to learn.

10. A trained model generated according to the training method of claim 9.

Images