Crop information estimation device, crop information estimation system, and crop information estimation program

Abstract

To estimate various information on a crop with high accuracy.SOLUTION: A crop information estimation device includes a generation part for generating an estimation model for estimating three-dimensional data from a two-dimensional image of seedlings with the two-dimensional image of the seedlings of a crop and the three-dimensional data of the seedlings corresponding to the two-dimensional image as teacher data, a three-dimensional data estimation part for estimating three-dimensional data of first seedlings by inputting a two-dimensional image of the first seedlings to the estimation model, and a seedling information estimation part for estimating the volume of the first seedlings on the basis of the three-dimensional data of the first seedlings, or estimating the volume of the first seedlings on the basis of the three-dimensional data of the first seedlings to estimate the weight of the first seedlings from the estimated volume of the first seedlings on the basis of a relation between the volume and the weight of seedlings of the crop.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present invention relates to a crop information estimation device, a crop information estimation system, and a crop information estimation program. [Background technology]

[0002] In response to recent abnormal weather, attention is being focused on artificial light plant factories, which can ensure a stable supply of agricultural products throughout the year. However, artificial light plant factories consume a huge amount of electricity for equipment such as lighting for growing crops, making GHG emissions (Scope 2) an issue. GHG emissions (Scope 2) refer to the greenhouse gas emissions emitted when electricity, heat, and steam supplied by other companies, such as electricity purchased from a power company, are used.

[0003] To solve this problem, producers need to develop optimal cultivation plans by comparing and considering (1) the reduction of GHG emissions (Scope 2) by selecting types and varieties of leafy vegetables (especially non-heading lettuce) suitable for artificial light plant factories and by reducing the number of cultivation days, and (2) the productivity of the products.

[0004] To achieve this, it is necessary to construct a growth model that estimates various information related to crop growth from environmental and biological information. [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Masago Moriyuki, "Study on lettuce growth prediction for improving productivity in plant factories," 2019,<URL:https: / / doi.org / 10.24729 / 00000260> Summary of the Invention [Problem to be solved by the invention]

[0006] The above-mentioned Non-Patent Document 1 describes estimating fresh weight using projected leaf area obtained from an image taken from above the head vegetable, and using the estimated fresh weight for growth prediction. However, projected leaf area does not include information in the height direction. Therefore, if information in the height direction could be reflected in some form, it is thought that the accuracy of fresh weight estimation and growth prediction could be further improved.

[0007] Therefore, an object of the present invention is to provide a crop information estimation device, a crop information estimation system, and a crop information estimation program that are capable of accurately estimating various types of information related to crops. [Means for solving the problem]

[0008] The crop information estimation device of the present invention includes a generation unit that generates an estimation model that estimates 3D data from the 2D image of a seedling of a crop using a 2D image of the seedling and 3D data of the seedling corresponding to the 2D image as training data; a 3D data estimation unit that estimates 3D data of a first seedling by inputting the 2D image of the first seedling into the estimation model; and a seedling information estimation unit that estimates the volume of the first seedling based on the 3D data of the first seedling, or estimates the volume of the first seedling based on the 3D data of the first seedling and estimates the weight of the first seedling from the estimated volume of the first seedling based on the relationship between the volume and weight of the crop seedling. [Effects of the Invention]

[0009] The crop information estimation device, the crop information estimation system, and the crop information estimation program of the present invention have the effect of being able to estimate various pieces of information about crops with high accuracy. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram schematically illustrating an artificial light-type plant factory according to one embodiment. [Figure 2] FIG. 2 is an enlarged view of one of the growing shelves. [Figure 3]FIG. 3 is a block diagram showing a schematic configuration of the environmental control system. [Figure 4] FIG. 4 is a diagram showing an example of the hardware configuration of the crop information estimation device. [Figure 5] FIG. 5 is a functional block diagram of the crop information estimation device. [Figure 6] FIG. 6 is a block diagram showing detailed functions of the cultivation result estimation unit. [Figure 7] Figure 7(a) is a diagram showing three-dimensional data of a seedling, Figure 7(b) is a diagram showing a two-dimensional image taken from a certain angle corresponding to Figure 7(a), and Figure 7(c) is a diagram showing the processing contents of the three-dimensional data estimation model learning unit. [Figure 8] FIG. 8 is a diagram showing the relationship between the volume and weight of seedlings. [Figure 9] FIG. 9 is a diagram showing the relationship between the time elapsed from the start of cultivation during the seedling raising period and the weight of the seedlings. [Figure 10] Figure 10(a) is a schematic diagram of a neural network as an example of a growth model, and Figure 10(b) is a diagram showing the relationship between the growth rate during seedling cultivation and the weight of the aboveground part 40 days after the start of cultivation. [Figure 11] FIG. 11 is a diagram showing the processing contents of the data conversion unit and the growth model learning unit. [Figure 12] FIG. 12 is a diagram showing the relationship between the above-ground part weight obtained as a result of actual cultivation and the estimated above-ground part weight. [Figure 13] FIG. 13 is a flowchart showing the processing of the crop information estimating device. DETAILED DESCRIPTION OF THE INVENTION

[0011] An artificial light-type plant factory 10 according to one embodiment will be described in detail below with reference to FIGS.

[0012] FIG. 1 is a diagram schematically illustrating an artificial light-type plant factory 10 according to this embodiment. The artificial light-type plant factory 10 is a factory for growing crops using artificial light. This artificial light-type plant factory 10 includes a building 12, and a number of growing shelves 14 are installed inside the building 12. Also, an environmental adjustment device 16 is provided inside the building 12 for adjusting the environment when raising and cultivating crops. Also, a sensor 18 (see FIG. 3) that measures environmental information is provided inside the building 12. The environmental adjustment device 16 controls the seedling-raising environment and the cultivation environment based on the measurement results of the sensor 18.

[0013] As an example, lettuce, spinach, and other leafy vegetable crops are grown on the growing shelves 14 of the artificial light-type plant factory 10. In this embodiment, the "seedling raising period" refers to the early stage of crop growth, a period from sowing until the crop reaches a certain weight (the seedling period). If the crop is to be planted, the seedling raising period refers to the period until planting. On the other hand, the "cultivation period" refers to the later stage of crop growth, a period from when the crop has grown to a certain weight until the crop is harvested. If the crop is to be planted, the cultivation period refers to the period after planting. Furthermore, "planting" refers to planting the crop in a place where the crop will be grown until harvest. In the artificial light-type plant factory 10, the planting density is appropriately changed during the seedling raising period and cultivation period to maximize the efficiency of light energy per plant.

[0014] FIG. 2 is an enlarged view of one of the growing shelves 14 in FIG. 1. As shown in FIG. 2, the growing shelf 14 has multiple (four in FIG. 2) support columns 22 and multiple shelves 24 arranged at intervals in the vertical direction along the support columns 22. A culture medium area and a cultivation tank for growing crops are provided on the upper surface of each shelf 24 except for the top shelf 24. The lower surfaces of each shelf 24 except for the bottom shelf 24 form a ceiling surface relative to the shelves 24 located below, and this surface is provided with multiple light sources, such as LEDs (light emitting diodes). The lower surfaces (ceiling surfaces) of each shelf 24 except for the bottom shelf 24 are provided with multiple cameras 26 as imaging devices for photographing the cultivated crops from above. The cameras 26 do not need to be fixed to the ceiling surface; for example, they may move along a moving lane provided on the ceiling surface. In this case, only one camera 26 may be provided on each ceiling surface, or multiple cameras 26 may be provided.

[0015] FIG. 3 is a block diagram showing a schematic configuration of an environmental control system 100 used to appropriately control the crop cultivation environment in the artificial light-type plant factory 10. As shown in FIG.

[0016] 3, the environmental control system 100 includes a camera 26, a sensor 18, and an environmental adjustment device 16 installed in the building 12 of the artificial light-type plant factory 10, an input device 32, a display device 34, and a crop information estimation device 30. The input device 32 is a device for a producer to input various information, and the display device 34 is a device for displaying estimation results and the like obtained by the crop information estimation device 30. The input device 32 and the display device 34 may be separate devices, or the input device 32 and the display device 34 may be integrated into one device, such as a smartphone or tablet terminal.

[0017] In this embodiment, the crop information estimation device 30 estimates the weight of the target seedling for each period (number of days or hours) elapsed since the start of cultivation (start of seedling raising) based on a two-dimensional image of the seedling (referred to as the target seedling) photographed by the camera 26 and information input by the producer into the input device 32.

[0018] FIG. 4 is a diagram illustrating an example of the hardware configuration of the crop information estimation device 30. As shown in FIG. 4, the crop information estimation device 30 includes a central processing unit (CPU) 90, a read-only memory (ROM) 92, a random access memory (RAM) 94, storage (a hard disk drive (HDD) or a solid state drive (SSD)) 96, an input / output interface 97, and a portable storage medium drive 99. These components of the crop information estimation device 30 are connected to a bus 98. In the crop information estimation device 30, the CPU 90 executes a program (including a crop information estimation program) stored in the ROM 92 or the storage 96, or a program read by the portable storage medium drive 99 from the portable storage medium 91, thereby realizing the functions of the components illustrated in FIG. 5. Note that the functions of the components illustrated in FIG. 5 may be realized by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

[0019] Figure 5 is a functional block diagram of the crop information estimation device 30. In the crop information estimation device 30, the CPU 90 executes a program to function as a model learning unit 40, an input information acquisition unit 50, a model selection unit 52, an image acquisition unit 54, a cultivation result estimation unit 56, and an optimal environment determination unit 58 as a control unit. Note that Figure 5 also illustrates a model storage unit 70 stored in a storage 96 or the like.

[0020] The model learning unit 40 learns (generates) models (a three-dimensional data estimation model and a growth model, which will be described later) used by the cultivation result estimation unit 56. The model learning unit 40 has a three-dimensional data estimation model learning unit 42 as a generation unit, a growth model learning unit 44, and a data conversion unit 46. Details of each unit of the model learning unit 40 will be described later.

[0021] The input information acquisition unit 50 acquires information input by the producer to the input device 32. The producer inputs information about the location of the crop cultivation (e.g., the facility name and latitude / longitude of the artificial light-type plant factory 10) and information about the type and variety of the crop being cultivated to the input device 32. The producer also inputs information about the future cultivation environment (temperature, CO2 concentration, hours of sunlight, PPFD, etc.). The information about the future cultivation environment may be one pattern or multiple patterns, but in this embodiment, multiple patterns are input as information about the future cultivation environment. The input information acquisition unit 50 transfers information necessary for model selection (e.g., cultivation location, type and variety) from the acquired information to the model selection unit 52, and transfers information necessary for estimating the cultivation result (e.g., information about the future cultivation environment) to the cultivation result estimation unit 56.

[0022] The model selection unit 52 selects a model to be used by the cultivation result estimation unit 56 from among the models stored in the model storage unit 70, based on the information acquired from the input information acquisition unit 50. Here, the model storage unit 70 has prepared 3D data estimation models and growth models for each location, crop type, and variety. Therefore, the model selection unit 52 selects a model corresponding to the information acquired from the input information acquisition unit 50, reads it out from the model storage unit 70, and passes it to the cultivation result estimation unit 56.

[0023] The image acquisition unit 54 acquires a two-dimensional image (in this embodiment, an RGB image taken from above) of the target seedling (first seedling) photographed by the camera 26. The image acquisition unit 54 passes the acquired image of the target seedling to the cultivation result estimation unit 56. The image acquisition unit 54 acquires multiple two-dimensional images taken while the target seedling is being grown. Each two-dimensional image is associated with the date and time it was photographed (information that can identify the period of time that has elapsed since the start of cultivation) and identification information for the seedling that was photographed.

[0024] The cultivation result estimation unit 56 estimates the weight of the target seedling for each period of time elapsed since the start of cultivation using multiple two-dimensional images of the target seedling during its cultivation, information on the future cultivation environment input by the producer, and the model selected by the model selection unit 52.

[0025] Here, the cultivation result estimation unit 56 has the functions shown in Fig. 6. Specifically, the cultivation result estimation unit 56 has the functions of a 3D data estimation unit 152 as an acquisition unit, a volume / weight prediction unit 154 as a seedling information estimation unit, a growth model coefficient calculation unit 156, and a growth estimation unit 158. The growth model coefficient calculation unit 156 and the growth estimation unit 158 ​​realize the function of a cultivation information estimation unit that estimates the weight of the crop for each period elapsed since the start of cultivation based on the time-series data of the seedling weight and information on the cultivation environment.

[0026] The 3D data estimation unit 152 uses a 3D data estimation model to estimate 3D data (point cloud data) of the seedling from a 2D image of the seedling captured by the camera 26. Here, the 3D data estimation model is a model that estimates and outputs 3D data of the seedling when a 2D image of the seedling is input. The 3D data estimation model is generated by the 3D data estimation model learning unit 42 of the model learning unit 40 shown in FIG. 5. Specifically, for the same type / variety of crop, many pairs of 3D data of the seedling as shown in FIG. 7(a) and corresponding sets of 2D images of the seedling (image sets including 2D images of the seedling photographed from one direction (above) as shown in FIG. 7(b), as well as 2D images of the seedling photographed from all angles (directions) different from FIG. 7(b) or from multiple angles (directions)) are prepared in advance. Then, as shown in Fig. 7(c), the 3D data estimation model learning unit 42 uses many sets of 3D data and 2D images of the seedlings as training data to learn a 3D estimation model that estimates the 3D data of the seedlings from the 2D images of the seedlings. This 3D data estimation model is generated for each item / variety and stored in the model storage unit 70 in a state where it is labeled by item / variety.

[0027] 6, the volume / weight prediction unit 154 predicts the volume of the target seedling based on the three-dimensional data of the target seedling estimated by the three-dimensional data estimation unit 152. Because the three-dimensional data is point cloud data, the volume / weight prediction unit 154 obtains the surface shape of the target seedling from the point cloud data and calculates the volume of the target seedling based on the surface shape.

[0028] Furthermore, the volume / weight prediction unit 154 predicts the weight of the seedlings based on the predicted volume of the seedlings. Here, it is assumed that the volume / weight prediction unit 154 holds the relationship between the volume and weight of the seedlings as shown in FIG. 8 for each item / variety. Note that the relationship in FIG. 8 is obtained as a result of actually measuring the volume and weight, and it is assumed that the volume / weight prediction unit 154 holds the formula of the approximate straight line shown by the dashed line for each item / variety. Therefore, the volume / weight prediction unit 154 predicts the weight of the target seedlings by substituting the predicted volume of the target seedlings into the formula of the approximate straight line. Note that the volume / weight prediction unit 154 can obtain time-varying data on the weight of the target seedlings from multiple three-dimensional data sets of the target seedlings during their raising.

[0029] The growth model coefficient calculation unit 156 calculates the value of the growth model coefficient using the time-lapse data of the weight of the target seedling (weight data predicted from multiple 2D images). It is known that the weight of a seedling increases exponentially with the elapsed time from the start of cultivation, as shown in Figure 9. Therefore, the growth model coefficient calculation unit 156 uses the time-lapse data of the weight of the target seedling to find an exponential function that indicates the change in weight with the elapsed time from the start of cultivation, and sets the value of the coefficient included in this exponential function as the value of the growth model coefficient.

[0030] The growth estimation unit 158 ​​inputs the values ​​of the growth model coefficients calculated by the growth model coefficient calculation unit 156 and (one pattern of) future environmental information input by the producer into the growth model, thereby estimating the weight of the target seedling for each period elapsed since the start of cultivation (for example, the weight on the 20th day, the weight on the 30th day, the weight on the 40th day, etc.).

[0031] 10(a), the growth model is assumed to be a neural network that can output the coefficient values ​​of a calculation formula used to calculate the weight for each period elapsed since the start of cultivation when the growth model coefficient values ​​and environmental information are input. The growth estimation unit 158 ​​substitutes the output coefficient values ​​into the calculation formula and calculates (estimates) the weight for each period elapsed since the start of cultivation using the calculation formula.

[0032] In addition, the inventors conducted experiments and found that, for example, in the case of lettuce seedlings, there is a relationship as shown in FIG. 10(b) between the growth rate during seedling raising and the aboveground weight after a predetermined period of time has elapsed (for example, 40 days since the start of cultivation), and that there is a certain correlation. Therefore, it is believed that there is a correlation between the values ​​of the exponential function coefficients (growth model coefficients) in FIG. 9 and the coefficients of the calculation formula output from the growth model, and also that there is a correlation between environmental information and the coefficients of the calculation formula, so it can be said that a growth model like that shown in FIG. 10(a) is feasible. Note that it is possible to use other AI (Artificial Intelligence) technologies, machine learning, deep learning, etc., in addition to neural networks.

[0033] Here, the growth model in Fig. 10(a) is generated by the data conversion unit 46 and the growth model learning unit 44 in Fig. 5. Fig. 11 shows a schematic flow of the process when generating (learning) a growth model.

[0034] When generating (learning) a growth model, first, a large number of data sets containing time-lapse data on the weight of actually cultivated seedlings, environmental information when the seedlings were actually cultivated, and weights of the actually cultivated seedlings (crops) for each period elapsed since the start of cultivation are prepared by labeling them with set values ​​(location, item / variety, etc.), and used as training data. Next, the data conversion unit 46 converts the time-lapse data on the weight of the seedlings included in each training data into growth model coefficients. The processing content of the data conversion unit 46 is the same as that of the growth model coefficient calculation unit 156 described above.

[0035] Next, the growth model learning unit 44 learns a growth model using the teacher data that has been partially converted by the data conversion unit 46. The growth model is generated for each set value (location, item / variety, etc.) and stored in the model storage unit 70 in a state where it is labeled by the set value.

[0036] The growth estimation unit 158 ​​estimates the weight of the target seedling for each elapsed period from the start of cultivation using each of multiple patterns of future environmental information input by the producer. The growth estimation unit 158 ​​transmits the estimation results to the optimal environment determination unit 58 in FIG. 5.

[0037] Here, the inventors actually cultivated lettuce and estimated the above-ground weight 40 days after the start of cultivation from two-dimensional images of the seedlings using the method described above. Then, they plotted a graph showing the relationship between the actual measured above-ground weight 40 days after the start of cultivation and the estimated above-ground weight. The results are shown in FIG. 12. As shown in FIG. 12, the approximation line is approximately y = x, and it was found that the weight for each period after the start of cultivation can be estimated with high accuracy.

[0038] Returning to FIG. 5 , the optimal environment determination unit 58 evaluates the estimated weight of the target seedlings for each period elapsed since the start of cultivation for each of the multiple patterns of future environmental information input by the producer, and determines the optimal pattern of future environmental information (optimal cultivation environment). At this time, the optimal environment determination unit 58 can determine the optimal cultivation environment from various perspectives. For example, the optimal environment determination unit 58 may determine an environment that can reduce GHG emissions (Scope 2) without significantly reducing the weight as the optimal cultivation environment. Furthermore, the optimal environment determination unit 58 may determine an environment that maximizes profits based on the utility costs spent in the number of days until harvest and the revenue calculated from the weight at harvest as the optimal cultivation environment.

[0039] The optimum environment determining unit 58 controls the environment adjusting device 16 while monitoring the measurement values ​​of the sensor 18 so as to achieve the determined optimum cultivation environment.

[0040] In addition, the optimal environment determination unit 58 may display on the display device 34 the change in weight for each period of time since the start of cultivation for each future cultivation environment pattern, allowing the producer to select which cultivation environment pattern to adopt.

[0041] (About processing) 13 is a flowchart showing the processing of the crop information estimating device 30. The flow of the processing of the crop information estimating device 30 will be described below with reference to FIG.

[0042] When the processing of FIG. 13 starts, first, in step S10, the input information acquisition unit 50 acquires input information (location, item / variety, and multiple patterns of future cultivation environment) entered into the input device 32, and the image acquisition unit 54 acquires multiple 2D images of the target seedling during the seedling raising period. Note that in this explanation, for simplicity, it is assumed that there is one target seedling, but this is not limiting, and there may be multiple target seedlings. In this case, the processing of FIG. 13 may be repeated as many times as the number of target seedlings.

[0043] Next, in step S12, the model selection unit 52 selects a model (three-dimensional data estimation model, growth model) to be used by the cultivation result estimation unit 56 from the models stored in the model storage unit 70 based on the input information (location, item / variety, etc.), and passes it to the cultivation result estimation unit 56.

[0044] Next, in step S14, the three-dimensional data estimation unit 152 of the cultivation result estimation unit 56 selects one of the acquired two-dimensional images that has not been selected. Next, in step S16, the three-dimensional data estimation unit 152 estimates three-dimensional data from the selected two-dimensional image using a three-dimensional data estimation model.

[0045] Next, in step S18, the volume / weight prediction unit 154 predicts the volume of the target seedling from the estimated three-dimensional data, and predicts the weight of the target seedling from the predicted volume and the volume-weight relationship as shown in Figure 8.

[0046] Next, in step S20, the volume / weight prediction unit 154 determines whether weights have been predicted from all of the acquired two-dimensional images. If this determination is negative, the process returns to step S14, and the process of predicting the weight of the target seedling from unselected two-dimensional images (S14, S16, S18) is executed. Thereafter, when the weight prediction process for all of the two-dimensional images has been completed and the determination in step S20 is positive, the process proceeds to step S22.

[0047] In step S22, the growth model coefficient calculation unit 156 calculates the value of the growth model coefficient (coefficient included in the exponential function as shown in FIG. 9) from the time-course data of the weight of the target seedling.

[0048] Next, in step S24, the growth estimation unit 158 ​​selects one pattern of the cultivation environment acquired by the input information acquisition unit 50 in step S10.

[0049] Next, in step S26, the growth estimation unit 158 ​​inputs the selected cultivation environment pattern and the calculated growth model coefficient values ​​into the growth model, and obtains the coefficient values ​​of the formula for calculating weight for each period elapsed since the start of cultivation.

[0050] Next, in step S28, the growth estimation unit 158 ​​estimates the weight of the target seedling for each period elapsed from the start of cultivation when the target seedling is cultivated in the selected cultivation environment pattern.

[0051] Next, in step S30, the growth estimation unit 158 ​​determines whether all input cultivation environment patterns have been selected. If the determination in step S30 is negative, the process returns to step S24, and processes using unselected cultivation environment patterns (S24, S26, S28) are executed. Thereafter, when the determination in step S30 is positive after the processes using all cultivation environment patterns have been completed, the process proceeds to step S32.

[0052] When the process proceeds to step S32, the optimal environment determination unit 58 identifies an optimal cultivation environment pattern based on the processing results of the growth estimation unit 158, displays the identified optimal cultivation environment pattern, and controls the environmental adjustment device 16 to perform environmental control to achieve the identified cultivation environment.

[0053] This completes the processing in FIG.

[0054] As described above in detail, in this embodiment, the 3D data estimation model learning unit 42 generates a 3D data estimation model that estimates 3D data of a seedling from a 2D image of the seedling, using a 2D image of the seedling and 3D data of the seedling corresponding to the 2D image as training data. The 3D data estimation unit 152 of the cultivation result estimation unit 56 then inputs a 2D image of the target seedling into the 3D data estimation model to estimate 3D data of the target seedling. The volume / weight prediction unit 154 estimates the volume of the target seedling based on the estimated 3D data and estimates the weight from the volume of the target seedling based on the relationship between volume and weight as shown in FIG. 8 . In this way, this embodiment estimates 3D data from a 2D image of the target seedling and then estimates the volume and weight of the target seedling from the 3D data, thereby enabling accurate estimation of the volume and weight of the target seedling. For example, compared to estimating weight from the seedling's projected leaf area, it is possible to estimate weight taking into account information about the seedling's height. Furthermore, in this embodiment, even when growing crops in a narrow space such as the growing shelf 14, by capturing a two-dimensional image from one direction (for example, from above), three-dimensional data can be estimated from the two-dimensional image, and the volume and weight can be predicted from the estimated three-dimensional data. This makes it possible to accurately predict the volume and weight of seedlings even in a narrow space such as the growing shelf 14 where three-dimensional data of the seedlings' shapes cannot be obtained using a 3D scanner or the like.

[0055] Furthermore, in this embodiment, the growth estimation unit 158 ​​estimates the weight of the target seedling for each period elapsed since the start of cultivation based on the historical weight data of the target seedling and environmental information for the future cultivation period. As a result, the weight for each period elapsed since the start of cultivation is estimated based on the historical weight data estimated, and therefore it is possible to estimate the weight for each period elapsed since the start of cultivation with greater accuracy than when, for example, the projected leaf area of ​​the seedling is used to estimate the weight for each period elapsed since the start of cultivation.

[0056] In addition, in this embodiment, the weight for each period elapsed since the start of cultivation is estimated for each future pattern of the cultivation environment, and the optimal pattern of the cultivation environment is identified based on the estimation results, and the cultivation environment is controlled. This makes it possible to control the cultivation environment to an appropriate one.

[0057] In the above embodiment, the model learning unit 40 in Fig. 5 is included in the crop information estimation device 30, but the present invention is not limited to this. For example, the model learning unit 40 may be included in an information processing device different from the crop information estimation device 30 (for example, a server connected to the crop information estimation device 30 via a network). In this case, the models may be collectively managed in the information processing device, and the crop information estimation device 30 may acquire and use the required models from the information processing device as appropriate.

[0058] In the above embodiment, in FIG. 6, the case where the volume / weight prediction unit 154 predicts the volume and weight of the seedlings has been described, but the present invention is not limited to this. For example, the volume / weight prediction unit 154 may predict only the volume of the seedlings. In this case, the growth model coefficient calculation unit 156 may calculate the growth model coefficients based on the data on the volume of the seedlings over time. Because there is a correlation between the volume and weight of the seedlings (FIG. 8), the change in volume can also be expressed by an exponential function as shown in FIG. 9. Therefore, the growth model coefficients can be calculated using the data on the volume of the seedlings over time.

[0059] In the above embodiment, the 2D image of the target seedling to be input to the 3D data estimation model is described as an image of the seedling photographed from above, but this is not limited to this. For example, the 2D image may be an image of the seedling photographed from the side, or may be an image photographed from another direction. The above-mentioned 3D data estimation model is trained using images of the seedling photographed from all angles or from multiple angles as training data, so that the 3D data of the seedling can be accurately estimated even when using 2D images other than 2D images photographed from above.

[0060] In the above embodiment, for example, when three-dimensional data of the target seedling can be acquired in the artificial light-type plant factory 10, the three-dimensional data may be acquired instead of the two-dimensional image. In this case, the three-dimensional data estimation unit 152 in FIG. 6 can be omitted.

[0061] In the above embodiment, the cultivation result estimation unit 56 estimates the weight of crops cultivated in the artificial light-type plant factory 10 for each period elapsed since the start of cultivation, but the present invention is not limited to this. For example, the cultivation result estimation unit 56 may estimate the weight of crops cultivated outdoors for each period elapsed since the start of cultivation. In this case, predicted values ​​of environmental information provided by the Japan Meteorological Agency or the like may be used as the pattern of future environmental information.

[0062] In the above embodiment, we have described a case where the weight of leafy vegetable seedlings is estimated for each period of time since the start of cultivation from two-dimensional images or three-dimensional data of the seedlings, but this is not limited to this. It is also possible to estimate the weight (i.e., size) of all fruit vegetable seedlings for each period of time since the start of cultivation from two-dimensional images or three-dimensional data of seedlings of other crops (e.g., fruit vegetables).

[0063] The above-described embodiment is a preferred example of the present invention, but the present invention is not limited to this and can be modified in various ways without departing from the spirit of the present invention. [Explanation of symbols]

[0064] 10 Artificial light plant factory 14 Growing Shelf 26 Camera (photographic device) 30 Crop information estimation device 42 3D data estimation model learning unit (generation unit) 58 Optimal environment determination unit (control unit) 152 3D data estimation unit (acquisition unit) 154 Volume / Weight Prediction Unit (Seedling Information Estimation Unit) 156 Growth model coefficient calculation unit (part of cultivation information estimation unit) 158 Growth Estimation Department (part of the Cultivation Information Estimation Department)

Claims

1. a generation unit that generates an estimation model for estimating three-dimensional data from a two-dimensional image of a seedling of a crop, using a two-dimensional image of the seedling and three-dimensional data of the seedling corresponding to the two-dimensional image as training data; a three-dimensional data estimation unit that estimates three-dimensional data of the first seedling by inputting a two-dimensional image of the first seedling into the estimation model; a seedling information estimation unit that estimates a volume of the first seedling based on the three-dimensional data of the first seedling, or that estimates a volume of the first seedling based on the three-dimensional data of the first seedling and estimates a weight of the first seedling from the estimated volume of the first seedling based on a relationship between the volume and weight of the seedling of the crop; A crop information estimation device comprising:

2. 2. The crop information estimation device according to claim 1, further comprising a cultivation information estimation unit that estimates the weight of the first seedling for each elapsed period from the start of cultivation if the first seedling were cultivated based on time-series data on the volume or weight of the first seedling and information on the cultivation environment of the first seedling.

3. The crop information estimation device according to claim 2, characterized in that the cultivation information estimation unit estimates the weight of the first seedling for each elapsed period from the start of cultivation when the first seedling is cultivated in each cultivation environment by varying information regarding the cultivation environment of the first seedling, and evaluates each cultivation environment based on the estimation results.

4. The crop information estimating device according to claim 3 , further comprising a control unit that controls the cultivation environment based on the evaluation result of the cultivation information estimating unit.

5. The crop information estimation device of claim 2, characterized in that the cultivation information estimation unit estimates the weight of the first seedling for each period elapsed from the start of cultivation using a growth model generated using training data including time-course data on the volume or weight of the crop seedling, information on the cultivation environment when the seedling was cultivated, and the weight of the seedling for each period elapsed from the start of cultivation when the seedling was cultivated.

6. The crop information estimating device according to claim 1 , wherein the crop is a leafy vegetable.

7. an acquisition unit that acquires three-dimensional data of a first seedling of a crop; a seedling information estimation unit that estimates a volume of the first seedling based on the three-dimensional data of the first seedling, or that estimates a volume of the first seedling based on the three-dimensional data of the first seedling and estimates a weight of the first seedling from the estimated volume of the first seedling based on a relationship between the volume and weight of the seedling of the crop; a cultivation information estimation unit that estimates a weight of the first seedling for each elapsed period from the start of cultivation based on time-varying data of the volume or weight of the first seedling and information on the cultivation environment of the first seedling; A crop information estimation device comprising:

8. an imaging device that captures a two-dimensional image of a crop seedling; A crop information estimating system comprising: the crop information estimating device according to any one of claims 1 to 6, which uses two-dimensional images captured by the imaging device.

9. The seedlings of the crop are cultivated on a growing shelf in a plant factory, 9. The crop information estimating system according to claim 8, wherein the photographing device photographs the seedlings being cultivated on the growing shelves.

10. generating an estimation model for estimating three-dimensional data from a two-dimensional image of a seedling of a crop using a two-dimensional image of the seedling and three-dimensional data of the seedling corresponding to the two-dimensional image as training data; a two-dimensional image of a first seedling is input to the estimation model to estimate three-dimensional data of the first seedling; Estimating a volume of the first seedling based on the three-dimensional data of the first seedling, or estimating a volume of the first seedling based on the three-dimensional data of the first seedling and estimating a weight of the first seedling from the estimated volume of the first seedling based on a relationship between the volume and weight of the seedling of the crop; A crop information estimation program that causes a computer to execute processing.

11. obtaining three-dimensional data of a first seedling of the crop; estimating a volume of the first seedling based on the three-dimensional data of the first seedling, or estimating a volume of the first seedling based on the three-dimensional data of the first seedling and estimating a weight of the first seedling from the estimated volume of the first seedling based on a relationship between the volume and weight of the seedling of the crop; Based on the data on the volume or weight of the first seedling over time and information on the cultivation environment of the first seedling, the weight of the first seedling is estimated for each elapsed period from the start of cultivation when the first seedling is cultivated. A crop information estimation program that causes a computer to execute processing.

Images